以下，邊參照圖面邊說明有關一實施形態。另外，各方塊圖的箭號是表示資料的流程。在本實施形態說明的各處理部，亦即機能區塊是可分別以不同的硬體所構成，亦可設為以1個的硬體共用複數的裝置或處理部的構成。又，雖詳細未圖示，但實際構成各裝置及各處理部的硬體是可分別將例如具有CPU、ROM、RAM及可改寫的快閃記憶體等的記憶區域之微電腦構成主體。記憶區域是記憶用以實現各裝置及各處理部的個體識別程式。然後，藉由在CPU中實行個體識別程式，可實現各裝置及各處理部的各處理。Hereinafter, one embodiment will be described with reference to the drawings. In addition, arrows in each block diagram indicate the flow of data. Each of the processing units described in this embodiment, that is, the functional blocks, may be configured by different hardwares, or may be configured to share a plurality of devices or processing units by one hardware. In addition, although not shown in detail, the actual hardware constituting each device and each processing unit can be constituted mainly by a microcomputer having memory areas such as CPU, ROM, RAM, and rewritable flash memory. The memory area stores individual identification programs for realizing each device and each processing unit. Then, by executing the individual identification program in the CPU, each process of each device and each processing unit can be realized.
亦即，在本實施形態說明的各處理部是藉由CPU實行被儲存於上述的記憶區域等的非遷移的實體的記憶媒體的電腦程式而實行對應於電腦程式的處理來實現，亦即藉由軟體來實現。另外，亦可設為藉由硬體來實現各處理部之中至少一部分來實現的構成。That is, each processing unit described in this embodiment is realized by the CPU executing the computer program stored in the non-migrating physical storage medium such as the above-mentioned memory area to execute processing corresponding to the computer program, that is, by using the computer program. implemented by software. In addition, a configuration in which at least a part of each processing unit is realized by hardware may be employed.
本實施形態的個體識別系統1是可比較預先被登錄的複數的登錄畫像與在攝影機器攝影的識別畫像，可從複數的登錄畫像之中識別拍攝有識別畫像的被照體的登錄畫像之系統。此個體識別系統1是例如可用在工廠等的零件的追蹤(trace)，但不被限於此。The individual identification system 1 of the present embodiment is a system that can compare a plurality of registered images registered in advance with the identification images captured by the camera, and can identify the registered image of the subject in which the identification image was captured from among the plurality of registered images. . This individual identification system 1 is, for example, usable in a trace of parts in a factory or the like, but is not limited thereto.
個體識別系統1是如圖1所示般，具備攝影機器10、輸入裝置11、顯示裝置12、個體識別裝置20及調整裝置30。本實施形態的情況，個體識別系統1是具備登錄側的攝影機器10及識別側的攝影機器10。在以下的說明中，區別該等時，將登錄側的攝影機器10稱為登錄側攝影機器10a，將識別側的攝影機器10稱為識別側攝影機器10b。在本實施形態中，登錄側攝影機器10a是攝取用以登錄於個體識別裝置20的畫像。又，識別側攝影機器10b是用以攝取所欲識別的被照體者。而且，在本實施形態中，所謂登錄畫像，主要是意思在登錄側的攝影機器10a攝影而被登錄於個體識別裝置20的畫像。又，所謂識別畫像，主要是意思在識別側的攝影機器10b攝影的畫像。As shown in FIG. 1 , the individual identification system 1 includes a camera 10 , an input device 11 , a display device 12 , an individual identification device 20 , and an adjustment device 30 . In the case of the present embodiment, the individual identification system 1 includes a camera device 10 on the registration side and a camera device 10 on the identification side. In the following description, when distinguishing these, the imaging device 10 on the registration side is referred to as the registration-side imaging device 10a, and the imaging device 10 on the recognition side is referred to as the recognition-side imaging device 10b. In the present embodiment, the registration-side camera 10 a captures an image for registration in the individual identification device 20 . In addition, the recognition-side photographing device 10b is used to capture a subject to be recognized. In addition, in the present embodiment, the "registered image" mainly means an image captured by the camera 10 a on the registration side and registered in the individual identification device 20 . In addition, the so-called identification image mainly means an image captured by the imaging device 10b on the identification side.
攝影機器10是具有照相機101、照明102及平台103而構成。照相機101是具有攝取成為工件等的攝影對象的被照體之機能。照相機101是被構成可依照來自外部的指示，變更快門速度或影像感測器的感度、解像度、攝影範圍、焦點等。照明102是用以照明被照體者，被構成可依照來自外部的指示，變更亮度或光度、色彩等。然後，平台103是支撐照相機101或照明102，且載置被照體者，被構成可變更照相機101或照明102與被照體的距離或角度亦即姿勢。The photographing apparatus 10 includes a camera 101 , an illumination 102 , and a platform 103 . The camera 101 has a function of capturing a subject that is an object of photography such as a workpiece. The camera 101 is configured to change the shutter speed, the sensitivity, resolution, imaging range, focus, and the like of the image sensor in accordance with an instruction from the outside. The illumination 102 is used for illuminating the object to be illuminated, and is configured to be able to change the brightness, luminosity, color, and the like in accordance with an instruction from the outside. Then, the platform 103 supports the camera 101 or the illumination 102 and places the subject, and is configured to change the distance or angle between the camera 101 or the illumination 102 and the subject, that is, the posture.
輸入裝置11是受理使用者的操作的介面，例如鍵盤或滑鼠等。使用者是可操作輸入裝置11來對於調整裝置30或個體識別裝置20進行各種的設定等。顯示裝置12是可文字或畫像等的顯示的顯示器，可顯示從個體識別裝置20或調整裝置30接收的資訊。使用者是藉由看被顯示於顯示裝置12的資訊，可確認個體識別裝置20或調整裝置30之處理的結果或設定內容等。The input device 11 is an interface that accepts user operations, such as a keyboard or a mouse. The user can operate the input device 11 to perform various settings and the like for the adjustment device 30 or the individual identification device 20 . The display device 12 is a display capable of displaying characters, images, and the like, and can display information received from the individual identification device 20 or the adjustment device 30 . By looking at the information displayed on the display device 12, the user can confirm the result of the processing by the individual identification device 20 or the adjustment device 30, the setting content, and the like.
個體識別裝置20是具有攝影裝置21、特徵抽出裝置22、登錄處理裝置23、記錄裝置24、字典裝置25、畫像對照識別裝置26、攝影條件賦予裝置27及特徵抽出條件賦予裝置28。本實施形態的情況，攝影裝置21、特徵抽出裝置22、登錄處理裝置23、記錄裝置24、攝影條件賦予裝置27及特徵抽出條件賦予裝置28是分別被設在登錄側及識別側。此情況，以虛線a所包圍的裝置成為登錄側的裝置，以虛線b所包圍的裝置成為識別側的裝置。另外，登錄側的裝置a及識別側的裝置b是亦可以不同的硬體所構成，或亦可以相同的硬體所構成。又，個體識別裝置20及調整裝置30是亦可被設置於被連接至網際網路等的電訊線路的雲端伺服器上。The individual identification device 20 includes a photographing device 21 , a feature extraction device 22 , a registration processing device 23 , a recording device 24 , a dictionary device 25 , an image matching identification device 26 , a photographing condition providing device 27 , and a feature extraction condition providing device 28 . In the present embodiment, the imaging device 21, the feature extraction device 22, the registration processing device 23, the recording device 24, the imaging condition providing device 27, and the feature extraction condition providing device 28 are provided on the registration side and the identification side, respectively. In this case, the device surrounded by the dotted line a is the device on the registration side, and the device surrounded by the dotted line b is the device on the recognition side. In addition, the device a on the registration side and the device b on the identification side may be constituted by different hardware, or may be constituted by the same hardware. In addition, the individual identification device 20 and the adjustment device 30 may be installed on a cloud server connected to a telecommunication line such as the Internet.
攝影裝置21是具有控制攝影機器10的驅動，攝取被照體的畫像而取得之機能。攝影裝置21是如圖2所示般，具有攝影處理部211及畫像取得處理部212。攝影處理部211是使攝影機器10以預定的條件驅動而攝取被照體。此情況，將攝影機器10所攝影時的預定的條件稱為攝影條件。攝影條件是包含例如關於攝影機器10的條件即攝影機器條件及關於成為攝影對象的被照體的條件即攝影對象條件。The photographing device 21 has a function of controlling the driving of the photographing device 10 to capture an image of the subject. As shown in FIG. 2 , the photographing device 21 includes a photographing processing unit 211 and an image acquiring processing unit 212 . The photographing processing unit 211 drives the photographing device 10 under predetermined conditions to photograph a subject. In this case, the predetermined conditions at the time of the photographing by the photographing device 10 are referred to as photographing conditions. The photographing conditions include, for example, photographing device conditions, which are conditions regarding the photographing device 10, and photographing target conditions, which are conditions regarding a subject to be photographed.
攝影機器條件，主要是關於光學機器的條件，例如有攝影機器10所具備的照相機101的快門速度、影像感測器的感度、解像度、攝影範圍、焦點、倍率、光圈、照明102的種類或亮度、光度、色彩等。攝影對象條件是關於攝影對象亦即被照體的條件，例如有被照體的姿勢，亦即被照體相對於照相機101或照明102的角度，照相機101或照明102與被照體的距離等。The camera conditions are mainly conditions related to optical equipment, such as the shutter speed of the camera 101 included in the camera 10, the sensitivity of the image sensor, the resolution, the shooting range, the focus, the magnification, the aperture, and the type or brightness of the lighting 102. , luminosity, color, etc. The photographic subject condition is the condition of the photographic subject, that is, the subject, such as the posture of the subject, that is, the angle of the subject relative to the camera 101 or the illumination 102, the distance between the camera 101 or the illumination 102 and the subject, etc. .
攝影處理部211是根據從攝影條件賦予裝置27接受的指示內容來改變攝影機器條件或攝影對象條件而使攝影機器10驅動，而可攝取被照體。The photographing processing unit 211 changes the photographing device condition or the photographing target condition according to the content of the instruction received from the photographing condition providing device 27 to drive the photographing device 10 to capture the subject.
畫像取得處理部212是可實行畫像取得處理。畫像取得處理是包含：取得在攝影機器10攝影的被照體的畫像而交給特徵抽出裝置22之處理。本實施形態的情況，登錄側a的畫像取得處理部212是以在攝影機器10a攝影的畫像作為登錄畫像交給登錄側a的特徵抽出裝置22。又，識別側b的畫像取得處理部212是以在攝影機器10b攝影的畫像作為識別畫像交給識別側b的特徵抽出裝置22。在本實施形態中，畫像取得處理之中，將在登錄側a進行之取得登錄畫像的處理稱為登錄畫像取得處理，將在識別側b進行之取得識別畫像的處理稱為識別畫像取得處理。另外，畫像取得處理是亦可包含：例如經由網際網路或經由記憶媒體而取得畫像等，從攝影機器10以外的手段取得畫像之處理。The image acquisition processing unit 212 is capable of executing image acquisition processing. The image acquisition process includes a process of acquiring an image of the subject photographed by the camera 10 and passing it to the feature extraction device 22 . In the case of the present embodiment, the image acquisition processing unit 212 of the registration side a uses the image captured by the camera 10a as the registration image to the feature extraction device 22 of the registration side a. In addition, the image acquisition processing unit 212 on the recognition side b uses the image captured by the photographing device 10b as a recognition image to hand over to the feature extraction device 22 on the recognition side b. In the present embodiment, among the image acquisition processes, the process of acquiring a registered image on the registration side a is called a registration image acquisition process, and the process of acquiring an identification image on the recognition side b is called a recognition image acquisition process. In addition, the image acquisition process may include, for example, a process of acquiring an image from means other than the camera 10, such as acquiring an image via the Internet or a storage medium.
特徵抽出裝置22是具有抽出從攝影裝置21取得的畫像中所含的特徵之機能。特徵抽出裝置22是具有計算特徵點、局部特徵量及廣域特徵量作為畫像的特徵之機能。特徵抽出裝置22是如圖3所示般，具有特徵點抽出處理部221、局部特徵量計算處理部222、局部特徵量群分類處理部223及廣域特徵量計算處理部224。特徵抽出裝置22是從特徵抽出條件賦予裝置28接受特徵抽出條件，實行各處理部221～224的處理。在特徵抽出條件中是含有例如抽出特徵點且計算局部特徵量時的用以決定金字塔畫像的階數及角落或似曲線的臨界值等。The feature extraction device 22 has a function of extracting features included in the image acquired from the photographing device 21 . The feature extraction device 22 has a function of calculating feature points, local feature amounts, and wide-area feature amounts as features of the image. As shown in FIG. 3 , the feature extraction device 22 includes a feature point extraction processing unit 221 , a local feature value calculation processing unit 222 , a local feature value cluster classification processing unit 223 , and a wide area feature value calculation processing unit 224 . The feature extraction device 22 receives the feature extraction conditions from the feature extraction condition setting device 28, and executes the processes of the respective processing units 221 to 224. The feature extraction conditions include, for example, threshold values for determining the order of the pyramid image and the corners or curve-like thresholds when extracting feature points and calculating local feature quantities.
特徵點抽出處理部221是可實行特徵點抽出處理。特徵點抽出處理是包含：如圖11所示般，從在畫像取得處理部212取得的畫像，抽出在被照體S的表面顯現的1個以上的特徵點P之處理。本實施形態的情況，特徵點抽出處理是從1張的畫像抽出例如數千點的特徵點。此情況，登錄側a的特徵點抽出處理部221是可抽出在登錄側的攝影機器10a攝影的登錄畫像的特徵點。又，識別側b的特徵點抽出處理部221是可抽出在識別側的攝影機器10b攝影的識別畫像的特徵點。The feature point extraction processing unit 221 is capable of executing feature point extraction processing. The feature point extraction process includes, as shown in FIG. 11 , a process of extracting one or more feature points P appearing on the surface of the subject S from the image acquired by the image acquisition processing unit 212 . In the case of the present embodiment, the feature point extraction process extracts, for example, several thousand feature points from a single image. In this case, the feature point extraction processing unit 221 on the registration side a can extract the feature points of the registration image captured by the camera 10a on the registration side. In addition, the feature point extraction processing unit 221 on the recognition side b is capable of extracting feature points of the recognition image captured by the camera 10b on the recognition side.
在此，例如圖12所示般，可思考藉由抽出顯現於工件S的表面的模樣之中角落或曲線等之類的非幾何學的模樣作為特徵點P，即使是在經同製造工程製造的同一種類的零件間也能識別該個體。這例如顯現於零件等的表面的直線或直角等的幾何學的模樣大多是在該種類的零件全體共通顯現的模樣，在零件間難以成為用以進行個體的識別的特徵。另一方面，可思考因為角落或曲線等的非幾何學的模樣大多是依照製造過程的極些微的條件的不同而形成者，該個體固有的情形多。因以，本實施形態的特徵點抽出處理部221是抽出在被照體的表面局部地顯現的角落或曲線的模樣作為特徵點P。Here, for example, as shown in FIG. 12 , it is conceivable to extract non-geometric patterns such as corners and curves among the patterns appearing on the surface of the workpiece S as the feature points P, even in the same manufacturing process. The individual can also be identified between parts of the same type. For example, geometric patterns such as straight lines and right angles appearing on the surface of parts and the like are often patterns that appear in common for all parts of this type, and are difficult to be used as features for individual identification among parts. On the other hand, it is thought that non-geometric shapes such as corners and curves are often formed according to the slightest conditions of the manufacturing process, and there are many cases that are unique to the individual. Therefore, the feature point extraction processing unit 221 of the present embodiment extracts, as the feature points P, corners or curves that appear locally on the surface of the subject.
本實施形態的特徵點抽出處理部221是利用從被照體的畫像取得的亮度梯度強度分佈，抽出局部地顯現於被照體的表面的角落或曲線作為特徵點。畫像是在畫素間具有亮度的梯度及其強度。均一的模樣時，亮度梯度強度的擴大小，又，直線的模樣時，亮度梯度強度是朝特定的方向擴大。另一方面，伴隨角落或曲線的模樣時，亮度梯度強度是朝複數方向擴大。因此，特徵點抽出處理部221是藉由尋求亮度梯度強度朝複數方向擴大的模樣，可抽出成為特徵點的角落或曲線。The feature point extraction processing unit 221 of the present embodiment extracts, as feature points, corners or curves that appear locally on the surface of the object by using the luminance gradient intensity distribution obtained from the image of the object. A portrait is a gradient of brightness and its intensity between pixels. In the case of a uniform pattern, the expansion of the intensity of the luminance gradient is small, and in the case of a straight pattern, the intensity of the luminance gradient is expanded in a specific direction. On the other hand, in the case of a corner or curve, the intensity of the luminance gradient increases in a complex number direction. Therefore, the feature point extraction processing unit 221 can extract corners or curves that become feature points by seeking a pattern in which the intensity of the luminance gradient expands in a complex number direction.
局部特徵量計算處理部222是可實行局部特徵量計算處理。局部特徵量計算處理是包含：根據包含在特徵點抽出處理被抽出的特徵點之周邊的區域的亮度梯度來計算特徵點的特徵量作為局部特徵量之處理。局部特徵量是具有特徵點的座標與256位元的二進位特徵量。在特徵量的計算是可使用既存技術的SIFT。又，若可能的話，亦可使用其他的SURF等的畫像特徵量計算算法。特徵點抽出處理是亦可求取將畫像多階段地縮小的金字塔畫像，對於各縮尺的畫像，以相同的條件來分別獨立地抽出特徵點。而且，局部特徵量計算處理部是亦可計算在各金字塔畫像取得的特徵點的局部特徵量。藉由如此，可求取具有規模(scale)不變性，不易受規模的不同影響的局部特徵量。The local feature value calculation processing unit 222 is capable of executing local feature value calculation processing. The local feature amount calculation process includes a process of calculating the feature amount of the feature point as the local feature amount based on the luminance gradient of the region included in the periphery of the feature point extracted by the feature point extraction process. The local feature amount is a binary feature amount having coordinates of feature points and 256 bits. The calculation of the feature quantity is SIFT which can use an existing technique. Also, if possible, other image feature value calculation algorithms such as SURF may be used. In the feature point extraction process, a pyramid image in which the image is reduced in multiple stages can be obtained, and feature points can be independently extracted for each of the reduced-scale images under the same conditions. Furthermore, the local feature amount calculation processing unit may also calculate the local feature amount of the feature points acquired in each pyramid image. By doing so, it is possible to obtain a local feature quantity that has scale invariance and is not easily affected by the difference in scale.
局部特徵量群分類處理部223是可實行局部特徵量群分類處理。局部特徵量群分類處理是成為用以計算登錄畫像及識別畫像的廣域特徵量的事前準備之處理。局部特徵量群分類處理是包含：例如圖13的(A)、(B)所示般，將針對各畫像實行局部特徵量計算處理而取得的複數的局部特徵量A予以按照該局部特徵量A的值來分類成預定數例如64～4096個的局部特徵量群B之處理。換言之，局部特徵量群分類處理是包含：將從各畫像取得的多數的局部特徵量A予以集合其值相似者彼此間而群集(clustering)的處理。構成局部特徵量群B的各個的要素Bn是具有至少1個以上從畫像取得的特徵點及其局部特徵量的一對。另外，Bn的n是表示要素的數量的整數。The local feature value group classification processing unit 223 is capable of executing the local feature value group classification process. The local feature value group classification process is a preparatory process for calculating the wide-area feature value of the registered image and identifying the image. The local feature value group classification process includes, for example, as shown in (A) and (B) of FIG. 13 , the complex number of local feature values A obtained by performing the local feature value calculation process for each image is assigned according to the local feature value A. A process of classifying the value of , into a predetermined number of, for example, 64 to 4096 local feature value groups B. In other words, the local feature value group classification process includes a process of clustering a large number of local feature values A obtained from each portrait by assembling those whose values are similar to each other. Each of the elements Bn constituting the local feature value group B is a pair having at least one feature point acquired from the image and its local feature value. In addition, n of Bn is an integer representing the number of elements.
局部特徵量群分類處理是包含：根據字典資訊d來分類從各畫像取得的局部特徵量A之處理。字典資訊d是從預先取得的學習用畫像所求得的特徵點的特徵量之中代表性者，在局部特徵量分類處理的實行前準備。局部特徵量群分類處理部223是將所欲分類的局部特徵量A充當成具有與該局部特徵量A接近的特徵量之字典資訊d而分類。The local feature value group classification process includes a process of classifying the local feature value A obtained from each portrait based on the dictionary information d. The dictionary information d is representative of the feature values of the feature points obtained from the learning images acquired in advance, and is prepared before the execution of the local feature value classification process. The local feature value group classification processing unit 223 classifies the local feature value A to be classified as dictionary information d having a feature value close to the local feature value A.
廣域特徵量計算處理部224是可實行廣域特徵量計算處理。廣域特徵量計算處理是包含：根據在局部特徵量分離處理取得的各局部特徵量群B來計算廣域特徵量之處理。廣域特徵量計算處理是包含：如圖13(C)所示般，以投票給與從畫像取得的各局部特徵量A相關性高的字典資訊d而得到的頻率分佈亦即柱狀圖作為廣域特徵量計算的處理。亦即，廣域特徵量是在橫軸取局部特徵量群B，在縱軸取投票給各局部特徵量群B的投票數。藉由利用如此計算的廣域特徵量，可用1個的特徵量來表示1張的畫像。廣域特徵量計算處理是例如可採用VLAD或BAG-OF-WORDS作為計算方法。The wide-area feature amount calculation processing unit 224 is capable of executing a wide-area feature amount calculation process. The wide-area feature amount calculation process includes a process of calculating a wide-area feature amount from each local feature amount group B obtained in the local feature amount separation process. As shown in FIG. 13(C), the wide-area feature amount calculation process includes, as shown in FIG. 13(C), a frequency distribution obtained by voting for dictionary information d having a high correlation with each local feature amount A obtained from the portrait, that is, a histogram. Processing of wide-area feature quantity calculation. That is, the wide-area feature values are the local feature value groups B on the horizontal axis, and the number of votes for each local feature value group B on the vertical axis. By using the wide-area feature amount calculated in this way, one image can be represented by one feature amount. In the wide-area feature amount calculation process, for example, VLAD or BAG-OF-WORDS can be used as a calculation method.
採用VLAD作為廣域特徵量的計算方法時，廣域特徵量計算處理是僅代表值的數量計算例如將局部特徵量與相關性高的字典資訊即代表值的差分向量總和後的向量V，以連接該等的向量作為廣域特徵量計算的向量V。具體而言，廣域特徵量計算處理部224是將從畫像得到的局部特徵量適用於其次的(1)式的計算而算出廣域特徵量的向量V。此情況，廣域特徵量的向量V是以64個的128次元的向量所構成。此情況，將各局部特徵量群B中所含的局部特徵量設為v，將字典資訊設為d。又，若將各局部特徵量群B中所含的特徵量的數量設為N，則成為1≦i≦N。
另外，(1)式的NN(v)=dk是意思在畫像中所含的特徵量之中，與此情況有64個的各字典資訊d相關性最高的局部特徵量v。When VLAD is used as the calculation method of the wide-area feature amount, the wide-area feature amount calculation process is to calculate only the number of representative values. These vectors are connected as a vector V for wide-area feature quantity calculation. Specifically, the wide-area feature amount calculation processing unit 224 calculates the vector V of the wide-area feature amount by applying the local feature amount obtained from the image to the calculation of the following equation (1). In this case, the vector V of the wide-area feature data is composed of 64 128-dimensional vectors. In this case, let the local feature value included in each local feature value group B be v, and the dictionary information be d. Furthermore, if the number of feature amounts included in each local feature amount group B is N, 1≦i≦N. In addition, NN(v)=dk in the formula (1) means that among the feature values included in the portrait, there are 64 local feature values v with the highest correlation with each dictionary information d in this case.
又，採用BAG-OF-WORDS作為廣域特徵量時，廣域特徵量字典產生處理是如其次般，計算廣域特徵量。此情況，廣域特徵量是持有字典資訊的數量的次元之向量，例如成為64次元的向量V。廣域特徵量字典產生處理部251是利用成為64個的代表值的字典資訊，針對各特徵量，尋找最近的字典資訊d，在接近的向量V的要素加上1。藉此，向量V是成為特徵量的柱狀圖，向量V的要素的總和是與特徵量的個數形成相等。此向量V是以此總數除算全體，作為正規化利用。Furthermore, when BAG-OF-WORDS is used as the wide-area feature amount, the wide-area feature amount dictionary generation process is the next step, and the wide-area feature amount is calculated. In this case, the wide-area feature amount is a dimensional vector holding the number of dictionary information, for example, a 64-dimensional vector V. The wide-area feature value dictionary generation processing unit 251 uses the dictionary information having 64 representative values to find the nearest dictionary information d for each feature value, and adds 1 to the element of the adjacent vector V. Thereby, the vector V is a histogram serving as a feature value, and the sum of the elements of the vector V is equal to the number of feature values. This vector V is used as normalization by dividing the whole by this total.
登錄處理裝置23是具有將在各處理取得的資料記錄於記錄裝置24之機能。登錄處理裝置23是如圖4所示般，具有檢索對象畫像登錄處理部231、評價用畫像登錄處理部232、局部特徵量登錄處理部233及廣域特徵量登錄處理部234。檢索對象畫像登錄處理部231是可實行檢索對象畫像登錄處理。檢索對象畫像登錄處理是包含：預先將作為複數的檢索對象的畫像設為登錄畫像，登錄於記錄裝置24的檢索對象畫像登錄部241之處理。The registration processing device 23 has a function of recording the data acquired in each process in the recording device 24 . As shown in FIG. 4 , the registration processing device 23 includes a retrieval target image registration processing unit 231 , an evaluation image registration processing unit 232 , a local feature value registration processing unit 233 , and a wide-area feature value registration processing unit 234 . The retrieval target image registration processing unit 231 is capable of executing retrieval target image registration processing. The retrieval-target image registration processing includes a process of preliminarily setting a plurality of retrieval-target images as registration images and registering them in the retrieval-target image registration unit 241 of the recording device 24 .
評價用畫像登錄處理部232是可實行評價用畫像登錄處理。評價用畫像登錄處理是包含：以在進行在畫像對照識別裝置26被實行的對照處理的評價時使用的畫像作為評價用的畫像登錄於記錄裝置24的評價用畫像登錄部242之處理。評價用畫像是構成包括：包含攝取了各個不同的被照體的複數的登錄畫像之登錄畫像群、及以和登錄畫像群之中的1個相同的被照體且不同的條件來攝取的識別畫像。此情況，識別畫像是亦可依被照體或條件的數量而為複數。又，攝取了相同的被照體的登錄畫像及識別畫像是例如藉由賦予相同或部分共通的識別號碼等等來互相建立關聯的狀態下登錄於記錄裝置24的評價用畫像登錄部242。The evaluation image registration processing unit 232 is capable of executing evaluation image registration processing. The image registration process for evaluation includes the process of registering the image used in the evaluation of the matching process executed by the image matching recognition device 26 as the image for evaluation in the evaluation image registration unit 242 of the recording device 24 . The image for evaluation consists of a registered image group including a plurality of registered images that have captured different subjects, and an identification image captured with the same subject and different conditions as one of the registered image groups. portrait. In this case, the identification images may be plural depending on the number of subjects or conditions. The registration image and the identification image that have captured the same subject are registered in the evaluation image registration unit 242 of the recording device 24 in a state in which they are associated with each other by, for example, assigning the same or partially common identification number or the like.
局部特徵量登錄處理部233是可實行局部特徵量登錄處理。局部特徵量登錄處理是包含：將藉由在特徵抽出裝置22的處理來從登錄畫像抽出的特徵點的局部特徵量登錄於記錄裝置24的局部特徵量登錄部243之處理。又，局部特徵量登錄處理是包含：將藉由在特徵抽出裝置22的處理來從評價用畫像抽出的特徵點的局部特徵量登錄於記錄裝置24的局部特徵量登錄部243之處理。局部特徵量登錄處理部233是將從各畫像取得的局部特徵量與成為抽出源的畫像建立關聯而登錄。The local feature value registration processing unit 233 is capable of executing local feature value registration processing. The local feature value registration process includes a process of registering the local feature value of the feature point extracted from the registered image by the processing in the feature extraction device 22 in the local feature value registration unit 243 of the recording device 24 . The local feature value registration process includes a process of registering the local feature value of the feature point extracted from the evaluation image by the processing in the feature extraction device 22 in the local feature value registration unit 243 of the recording device 24 . The local feature value registration processing unit 233 registers the local feature value acquired from each image in association with the image that is the source of extraction.
廣域特徵量登錄處理部234是可實行廣域特徵量登錄處理。廣域特徵量登錄處理是包含：將涉及登錄畫像的廣域特徵量登錄於記錄裝置24的廣域特徵量登錄部244之處理。又，廣域特徵量登錄處理是包含：將涉及評價用畫像的廣域特徵量登錄於記錄裝置24的廣域特徵量登錄部244之處理。局部特徵量登錄處理部233是將從各畫像取得的廣域特徵量與成為抽出源的畫像建立關聯而登錄。The wide-area feature value registration processing unit 234 is capable of executing a wide-area feature value registration process. The wide-area feature value registration process includes a process of registering the wide-area feature value related to the registered image in the wide-area feature value registration unit 244 of the recording device 24 . In addition, the wide-area feature value registration process includes a process of registering the wide-area feature value related to the image for evaluation in the wide-area feature value registration unit 244 of the recording device 24 . The local feature value registration processing unit 233 registers the wide-area feature value acquired from each image in association with the image that is the source of extraction.
記錄裝置24是具有記錄藉由登錄處理裝置23的各登錄處理所傳送的各種的資料之機能，構成資料庫。記錄裝置24是亦可例如構成個體識別裝置20的電腦的記錄裝置，或以經由網路線路來連接的外部的伺服器等所構成。記錄裝置24是如圖5所示般，具有檢索對象畫像登錄部241、評價用畫像登錄部242、局部特徵量登錄部243、廣域特徵量登錄部244及廣域特徵量字典登錄部245。The recording device 24 has a function of recording various data transmitted by each registration process of the registration processing device 23, and constitutes a database. The recording device 24 may be, for example, a recording device of a computer constituting the individual identification device 20, or an external server or the like connected via a network line. As shown in FIG. 5 , the recording device 24 includes a retrieval target image registration unit 241 , an evaluation image registration unit 242 , a local feature value registration unit 243 , a wide area feature value registration unit 244 , and a wide area feature value dictionary registration unit 245 .
檢索對象畫像登錄部241是記錄從登錄處理裝置23的檢索對象畫像登錄處理部231傳送的檢索對象畫像的記錄區域。評價用畫像登錄部242是記錄從登錄處理裝置23的評價用畫像登錄處理部232傳送的評價用畫像的記錄區域。局部特徵量登錄部243是記錄從登錄處理裝置23的局部特徵量登錄處理部233傳送的局部特徵量的記錄區域。廣域特徵量登錄部244是記錄從登錄處理裝置23的廣域特徵量登錄處理部234傳送的廣域特徵量的記錄區域。The search target portrait registration unit 241 is a recording area in which the search target portrait transferred from the search target portrait registration processing unit 231 of the registration processing device 23 is recorded. The evaluation image registration unit 242 is a recording area in which the evaluation image sent from the evaluation image registration processing unit 232 of the registration processing device 23 is recorded. The local feature value registration unit 243 is a recording area in which the local feature value transferred from the local feature value registration processing unit 233 of the registration processing device 23 is recorded. The wide area feature value registration unit 244 is a recording area for recording the wide area feature value transmitted from the wide area feature value registration processing unit 234 of the registration processing device 23 .
又，廣域特徵量字典登錄部245是記錄從字典裝置25的廣域特徵量字典登錄處理部252傳送的廣域特徵量字典的記錄區域。在畫像對照識別裝置26實行檢索處理以前，登錄有複數的廣域特徵量字典，該複數的廣域特徵量字典是使要素持有根據從預先取得的學習用畫像得到局部特徵量來計算的複數的字典資訊。In addition, the wide area feature value dictionary registration unit 245 is a recording area for recording the wide area feature value dictionary transmitted from the wide area feature value dictionary registration processing unit 252 of the dictionary device 25 . Before the image matching and recognizing device 26 executes the retrieval process, plural wide-area feature value dictionaries are registered, and the plural wide-area feature value dictionaries are plural numbers calculated based on the local feature values obtained from the preliminarily acquired learning images. dictionary information.
字典裝置25是以從複數張的畫像抽出的特徵點及其局部特徵量為基礎，產生複數的字典資訊，產生根據該字典資訊的廣域特徵量字典，具有登錄及設定的機能。字典裝置25是如圖6所示般，具有廣域特徵量字典產生處理部251及廣域特徵量字典登錄處理部252。The dictionary device 25 generates plural dictionary information based on feature points and local feature values extracted from a plurality of images, generates a wide-area feature value dictionary based on the dictionary information, and has functions of registration and setting. As shown in FIG. 6 , the dictionary device 25 includes a wide-area feature value dictionary generation processing unit 251 and a wide-area feature value dictionary registration processing unit 252 .
廣域特徵量字典產生處理部251是可實行廣域特徵量字典產生處理。廣域特徵量字典產生處理是包含：產生廣域特徵量字典之處理，該廣域特徵量字典是在字典資訊中持有例如從複數的學習用畫像取得的局部特徵量群的各個的代表值。此情況，廣域特徵量字典產生處理是包含：以局部特徵量群的重心作為代表值設定於字典資訊的處理。字典資訊是構成廣域特徵量字典的要素，集合複數個的字典資訊來構成1個的廣域特徵量字典。The wide-area feature amount dictionary generation processing unit 251 is capable of executing a wide-area feature amount dictionary generation process. The wide-area feature amount dictionary generation process includes: a process of generating a wide-area feature amount dictionary that holds, for example, a representative value of each local feature amount group obtained from a complex number of learning images in the dictionary information. . In this case, the wide-area feature amount dictionary generation process includes a process of setting the center of gravity of the local feature amount group as a representative value in the dictionary information. The dictionary information is an element constituting a wide-area feature quantity dictionary, and a plurality of dictionary information is assembled to form one wide-area feature quantity dictionary.
此情況，例如，在廣域特徵量字典產生處理中，廣域特徵量字典產生處理部251是可如其次般產生廣域特徵量字典。亦即，在廣域特徵量字典產生處理中，廣域特徵量字典產生處理部251是首先如圖14(A)、(B)所示般，接受例如在特徵抽出裝置22被計算的m張分的學習用畫像T1～Tm的局部特徵量A。其次，廣域特徵量字典產生處理部251是如圖14(C)所示般，例如利用k-means法來將接受的局部特徵量A予以分類成k個、此情況是64個的族群、亦即局部特徵量群B1～Bk，針對各局部特徵量群B1～Bk的各者，求取重心C1～Ck。然後，廣域特徵量字典產生處理部251是將如此求取的各局部特徵量群B1～Bk的重心C1～Ck予以設定於字典資訊d1～dk，作為各個的局部特徵量群B1～Bk的代表值，產生匯集字典資訊d1～dk的廣域特徵量字典E。然後，廣域特徵量字典產生處理部251是進一步利用不同的學習用畫像T來產生複數的廣域特徵量字典E。另外，字典資訊d1～dk只要是代表各局部特徵量群B1～Bk的值即可，不被限於重心。In this case, for example, in the wide-area feature amount dictionary generation process, the wide-area feature amount dictionary generation processing unit 251 can generate the wide-area feature amount dictionary as the next. That is, in the wide-area feature amount dictionary generation process, the wide-area feature amount dictionary generation processing unit 251 first receives m pieces of data calculated by, for example, the feature extraction device 22 as shown in FIGS. 14(A) and (B) . The local feature amounts A of the learning images T1 to Tm are divided into two parts. Next, as shown in FIG. 14(C), the wide-area feature value dictionary generation processing unit 251 classifies the received local feature values A into k, in this case, 64 groups, for example, using the k-means method. That is, for the local feature value groups B1 to Bk, the centers of gravity C1 to Ck are obtained for each of the local feature value groups B1 to Bk. Then, the wide-area feature value dictionary generation processing unit 251 sets the centers of gravity C1 to Ck of the respective local feature value groups B1 to Bk obtained in this way in the dictionary information d1 to dk, as the values of the respective local feature value groups B1 to Bk. As the representative value, a wide-area feature quantity dictionary E that collects dictionary information d1 to dk is generated. Then, the wide-area feature amount dictionary generation processing unit 251 further generates a complex wide-area feature amount dictionary E using the different learning images T. In addition, the dictionary information d1 to dk is not limited to the center of gravity as long as it is a value representing each of the local feature value groups B1 to Bk.
廣域特徵量字典登錄處理部252是可實行廣域特徵量字典登錄處理。廣域特徵量字典登錄處理是包含：將在廣域特徵量字典產生處理所產生的複數的廣域特徵量字典Eic記錄於記錄裝置24的廣域特徵量字典登錄部245之處理。此情況，在廣域特徵量字典E是含有構成廣域特徵量字典E的複數的字典資訊d1～dk。The wide-area feature amount dictionary registration processing unit 252 is capable of executing the wide-area feature amount dictionary registration process. The wide-area feature amount dictionary registration process includes a process of recording the plural wide-area feature amount dictionary Eic generated in the wide-area feature amount dictionary generation process in the wide area feature amount dictionary registration unit 245 of the recording device 24 . In this case, in the wide-area feature amount dictionary E, the dictionary information d1 to dk including complex numbers constituting the wide-area feature amount dictionary E are included.
畫像對照識別裝置26是具有：對照被記錄於記錄裝置24的登錄畫像與從識別側的特徵抽出裝置22輸出的識別畫像，識別該畫像的一致不一致之機能。畫像對照識別裝置26是如圖7所示般，具有檢索處理部261、對照處理部262及結果通知處理部263。The image collation and identification device 26 has a function of comparing the registered image recorded in the recording device 24 with the identification image output from the feature extraction device 22 on the identification side, and recognizing whether the images match or not. As shown in FIG. 7 , the image matching recognition device 26 includes a retrieval processing unit 261 , a matching processing unit 262 , and a result notification processing unit 263 .
檢索處理部261是可實行檢索處理。檢索處理是包含：檢索被記錄於記錄裝置24的檢索對象畫像登錄部241的涉及複數的登錄畫像的廣域特徵量之中，與從識別側的特徵抽出裝置22傳送的涉及識別畫像的廣域特徵量相關性高者之處理。本實施形態的情況，檢索處理部261是具有可實行鎖定處理的鎖定處理部2611及可實行特定處理的特定處理部2612。亦即，本實施形態的情況，檢索處理是包含鎖定處理及特定處理。The retrieval processing unit 261 is capable of executing retrieval processing. The retrieval process includes retrieving the wide-area feature values related to plural registered images recorded in the retrieval-target image registration unit 241 of the recording device 24, and the wide area related to the recognition images transmitted from the feature extraction device 22 on the recognition side. Processing of those with high correlation of feature quantities. In the case of the present embodiment, the search processing unit 261 includes a lock processing unit 2611 capable of executing lock processing and a specific processing unit 2612 capable of executing specific processing. That is, in the case of this embodiment, the retrieval process includes the lock process and the specific process.
在鎖定處理部2611被實行的鎖定處理是包含：如圖15所示般，從被登錄於記錄裝置24的檢索對象畫像登錄部241的複數的登錄畫像之中，以具有與識別畫像所具有的廣域特徵量的相關性高的廣域特徵量之預定數的登錄畫像作為候補鎖定的處理。當被登錄於檢索對象畫像登錄部241的登錄畫像例如為數千～數十萬張時，登錄畫像的候補是在鎖定處理被鎖定至例如數～數十張程度。The lock processing executed by the lock processing unit 2611 includes, as shown in FIG. 15 , from among a plurality of registered images registered in the retrieval-target image registration unit 241 of the recording device 24 , having the same image as that of the identification image. The process of locking a predetermined number of registered portraits of the wide-area feature amount having a high correlation of the wide-area feature amount as a candidate is locked. When the number of registered portraits registered in the search-target portrait registration unit 241 is, for example, several thousand to several hundreds of thousands, the candidates for the registered portraits are locked to, for example, several to several tens of pictures in the locking process.
在特定處理部2612被實行的特定處理是可利用：使用局部特徵量的畫像對照手法，及驗證特徵點的幾何性的配置的整合性的算法來進行。特定處理部2612是首先比較藉由鎖定處理所抽出的候補的登錄畫像的局部特徵量與識別畫像的局部特徵量，從雙方的畫像求取特徵量的差成為最小的特徵點作為一對。然後，特定處理部2612是抽出與其他的特徵點的相對性的位置關係不矛盾的特徵點作為對應點，特定該對應點數最多的登錄畫像。The specific processing performed by the specific processing unit 2612 can be performed by an image matching method using local feature amounts and an algorithm for verifying the consistency of the geometric arrangement of the feature points. The identification processing unit 2612 first compares the local feature value of the candidate registration image extracted by the locking process with the local feature value of the recognition image, and obtains a pair of feature points whose feature value difference is the smallest from the two images. Then, the identification processing unit 2612 extracts feature points whose relative positional relationship with other feature points does not contradict as corresponding points, and specifies the registered image with the largest number of corresponding points.
此情況，鎖定處理是如圖15所示般，比較登錄畫像與識別畫像的各個的廣域特徵量之處理，因此精度雖低於比較畫像中所含的全部的局部特徵量的特定處理，但處理量少可高速地處理。相對於此，特定處理是如圖16所示般，針對登錄畫像及識別畫像的各個的局部特徵量A的全部，探索對應點的處理，因此處理量及時間雖多於比較廣域特徵量彼此間的鎖定處理，但精度佳。如此，檢索處理是藉由實行鎖定處理與特定處理的2階段的處理，可高速且精度佳地從複數的登錄畫像之中特定與識別畫像相同的登錄畫像。In this case, as shown in FIG. 15, the locking process is a process of comparing the wide-area feature values of the registered portrait and the recognition portrait, so the accuracy is lower than the specific process of comparing all the local feature values included in the portrait, but High-speed processing is possible with a small amount of processing. On the other hand, as shown in FIG. 16 , the specific processing is a process of searching for corresponding points for all the local feature values A of the registration image and the recognition image, so the processing amount and time are larger than comparing the wide-area feature values with each other. Lock processing between time, but good accuracy. In this way, the retrieval process is performed in two stages of the locking process and the specifying process, and it is possible to specify the same registration image as the identification image from among a plurality of registration images at high speed and with high accuracy.
對照處理部262是可實行對照處理。對照處理是不伴隨畫像的檢索的處理，在為了藉由調整裝置30來調整各種的條件或算法、參數等而對照特定的畫像彼此間時被實行。對照處理是利用與檢索處理之中用以對照畫像彼此間的算法相同的算法來被實行。因此，對照處理部262是可作為檢索處理部261的一部分。藉由調整裝置30所調整的各種的條件或算法、參數等是被反映於檢索處理部261的檢索處理。The collation processing unit 262 is capable of executing collation processing. The matching process is a process that does not involve retrieval of images, and is executed when comparing specific images with each other in order to adjust various conditions, algorithms, parameters, and the like by the adjustment device 30 . The collation process is performed using the same algorithm as the algorithm used to collate the portraits with each other in the retrieval process. Therefore, the collation processing unit 262 can be used as a part of the retrieval processing unit 261 . Various conditions, algorithms, parameters, and the like adjusted by the adjustment device 30 are reflected in the retrieval processing of the retrieval processing unit 261 .
對照處理是包含：比較例如作為評價用畫像被登錄於評價用畫像登錄部242的特定的登錄畫像與識別畫像的局部特徵量，取得該特徵點的對應點數來對照登錄畫像與識別畫像的處理。此情況，對照處理部262是可依據來自調整裝置30的指示，針對攝取了相同的被照體的登錄畫像及識別畫像進行對照。又，對照處理部262是亦可依據來自調整裝置30的指示，針對攝取了不同的被照體的登錄畫像及識別畫像進行對照。對照處理部262是輸出登錄畫像與識別畫像的局部特徵量的對應點數。The collation process includes, for example, a process of comparing the local feature amount of a specific registration image registered in the evaluation image registration unit 242 as an evaluation image registration unit 242 with the recognition image, and obtaining the corresponding points of the feature points to compare the registration image and the recognition image. . In this case, the collation processing unit 262 can compare the registered image and the identification image in which the same subject has been captured in accordance with an instruction from the adjustment device 30 . In addition, the collation processing unit 262 can also collate the registered image and the identification image of the different subjects captured according to an instruction from the adjustment device 30 . The matching processing unit 262 outputs the number of corresponding points between the registered image and the local feature value of the recognized image.
又，對照處理部262是可利用從對照識別條件調整裝置32接受的參數組來實行對照處理。參數組是例如圖17所示般，針對用以設定檢索處理及對照處理的條件之複數的參數，組合可設定於各個的參數組合的值者。圖17的條件1、條件2、條件3•••會分別成為1個的參數組。圖17的例子的情況，參數組是全部成為5400種。In addition, the matching processing unit 262 can execute the matching processing using the parameter group received from the matching identification condition adjustment device 32 . The parameter group is, for example, as shown in FIG. 17 , for a plurality of parameters for setting the conditions of the retrieval processing and the matching processing, the values that can be set in each parameter combination are combined. Condition 1, Condition 2, and Condition 3••• in Figure 17 are each a parameter group of one. In the case of the example of FIG. 17 , there are 5400 types of parameter sets in total.
結果通知處理部263是可實行結果通知處理。結果通知處理是包含：將檢索處理的結果及對照處理的結果輸出至調整裝置30或外部的裝置之處理。此情況，檢索處理的結果是含有關於特定的登錄畫像的資訊。又，對照處理的結果是含有對應點數。The result notification processing unit 263 is capable of executing the result notification processing. The result notification process includes a process of outputting the result of the search process and the result of the comparison process to the adjustment device 30 or an external device. In this case, the result of the search process includes information about a specific login image. In addition, the result of the collation processing includes the corresponding number of points.
攝影條件賦予裝置27是具有：接受來自調整裝置30的攝影條件調整裝置31的指示，對攝影裝置21賦予上述的攝影條件之機能。特徵抽出條件賦予裝置28是具有：接受來自調整裝置30的對照識別條件調整裝置32的指示，對特徵抽出裝置22賦予特徵抽出條件之機能。The photographing condition providing device 27 has a function of applying the above-described photographing conditions to the photographing device 21 in response to an instruction from the photographing condition adjusting device 31 from the adjusting device 30 . The feature extraction condition applying means 28 has a function of applying a feature extraction condition to the feature extraction means 22 in response to an instruction from the matching recognition condition adjustment means 32 from the adjustment means 30 .
調整裝置30是具有將從畫像對照識別裝置26取得的結果可視化而顯示於顯示裝置12的機能，或接受來自畫像對照識別裝置26的反餽，調整攝影機器10的攝影條件或在特徵抽出裝置22使用的條件或算法等、在個體識別裝置20使用的各種的條件、算法、或參數等的機能等。調整裝置30是如圖1所示般，具有攝影條件調整裝置31、對照識別條件調整裝置32及可視化裝置33。The adjusting device 30 has a function of visualizing the result obtained from the image matching and recognizing device 26 and displaying it on the display device 12 , or receiving feedback from the image matching and identifying device 26 to adjust the shooting conditions of the camera 10 or use it in the feature extraction device 22 . conditions, algorithms, and the like, various conditions, algorithms, and functions of parameters and the like used in the individual identification device 20 . The adjustment device 30 includes, as shown in FIG. 1 , an imaging condition adjustment device 31 , a collation recognition condition adjustment device 32 , and a visualization device 33 .
攝影條件調整裝置31是具有：接受來自畫像對照識別裝置26的反餽而自動地，或接受來自使用者的操作而手動地，調整利用攝影機器10來攝取攝影對象時的各種條件，對於攝影條件賦予裝置27指示其條件的機能。使用者是例如藉由操作輸入裝置11，可依照攝影條件調整裝置31的處理來調整各種條件。又，攝影條件調整裝置31是具有按照藉由特徵抽出裝置22的處理所取得的特徵點的數量來自動調整各種條件的機能。攝影條件調整裝置31是如圖8所示般，具有攝影機器條件調整處理部311、攝影對象條件調整處理部312、總括條件產生處理部313及最適條件提示處理部314。The photographing condition adjusting device 31 is configured to automatically adjust various conditions when photographing a photographic subject by the photographing device 10 by automatically receiving feedback from the image matching and identifying unit 26 or manually by receiving an operation from the user, and assigning a photographing condition to the photographing condition adjusting device 31 . Device 27 indicates the function of its condition. For example, by operating the input device 11, the user can adjust various conditions in accordance with the processing of the photographing condition adjustment device 31. In addition, the imaging condition adjustment device 31 has a function of automatically adjusting various conditions according to the number of feature points acquired by the processing of the feature extraction device 22 . As shown in FIG. 8 , the photographing condition adjusting device 31 includes a photographing device condition adjusting processing unit 311 , a photographing target condition adjusting processing unit 312 , a collective condition generating processing unit 313 , and an optimum condition presentation processing unit 314 .
攝影機器條件調整處理部311可實行調整攝影機器條件的攝影機器條件調整處理。藉由攝影機器條件調整處理的實行，主要調整有關光學機器的條件，例如攝影機器10所具有的照相機101的快門速度或影像感測器的感度、解像度、攝影範圍、焦點、照明102的亮度或光度、色彩等。攝影機器條件調整處理部311是將被調整的攝影機器條件發送至攝影條件賦予裝置27。藉此，攝影裝置21是按照在攝影機器條件調整處理部311被調整的攝影機器條件來使攝影機器10的主要照相機101或照明102動作而攝取攝影對象。The camera condition adjustment processing unit 311 can execute camera condition adjustment processing for adjusting camera conditions. By executing the camera condition adjustment process, conditions related to optical equipment, such as the shutter speed of the camera 101 of the camera 10 or the sensitivity, resolution, photographing range, focus, brightness of the lighting 102 or the image sensor of the camera 10 are mainly adjusted. Luminosity, color, etc. The camera condition adjustment processing unit 311 transmits the adjusted camera conditions to the imaging condition setting device 27 . In this way, the photographing device 21 operates the main camera 101 or the lighting 102 of the photographing device 10 according to the photographing device conditions adjusted by the photographing device condition adjustment processing unit 311 to photograph the photographic subject.
攝影對象條件調整處理部312是可實行調整有關成為攝影對象的被照體的條件的攝影對象條件之攝影對象條件調整處理。藉由攝影對象條件調整處理的實行，例如調整被照體的姿勢，亦即被照體相對於照相機101或照明102的角度、照相機101或照明102與被照體的距離等。攝影對象條件調整處理部312是將被調整的攝影對象條件發送至攝影條件賦予裝置27。藉此，攝影裝置21是按照在攝影對象條件調整處理部312被調整的攝影對象條件來使攝影機器10的主要平台103動作而攝取攝影對象。The photographing target condition adjustment processing unit 312 is capable of executing photographing target condition adjustment processing that adjusts photographing target conditions related to conditions of a subject to be photographed. By executing the subject condition adjustment process, for example, the posture of the subject is adjusted, that is, the angle of the subject relative to the camera 101 or the illumination 102, the distance between the camera 101 or the illumination 102 and the subject, and the like. The photographing target condition adjustment processing unit 312 transmits the adjusted photographing target condition to the photographing condition providing device 27 . Thereby, the photographing device 21 operates the main stage 103 of the photographing apparatus 10 according to the photographing target condition adjusted by the photographing target condition adjustment processing unit 312 to photograph the photographing target.
總括條件產生處理部313是可實行總括條件產生處理。總括條件產生處理是包含：預先被設定的範圍內的攝影對象條件與預先被設定的範圍內的攝影機器條件的組合之處理。此情況，在攝影裝置21的畫像取得處理部212被實行的畫像取得處理是包含：取得以在總括條件產生處理產生的部的條件攝影後的畫像之處理。藉此，畫像取得處理部212是可取得以不同的條件攝影的多數的畫像，亦即可取得以攝影對象條件及攝影機器條件的各者來攝影後的多數的畫像。The collective condition generation processing unit 313 is capable of executing collective condition generation processing. The collective condition generation process is a process including a combination of the conditions of the photographing object within the range set in advance and the conditions of the photographing machine within the range set in advance. In this case, the image acquisition processing executed by the image acquisition processing unit 212 of the photographing device 21 includes the processing of acquiring the image captured under the conditions of the unit that generated the collective condition generation processing. Thereby, the image acquisition processing unit 212 can acquire a large number of images photographed under different conditions, that is, can acquire a large number of images that have been photographed under the conditions of the photographing object and the conditions of the photographing device.
例如圖19的例子，設定被照體與照明102的距離即照明距離，作為攝影對象條件，設定照明102的光量即光量設定值，作為攝影機器條件。此情況，總括條件產生處理部313是總括產生照明距離與光量設定值的取得的值的全部的組合。然後，畫像取得處理部212是以此總括的組合的條件，自動攝影，而取得登錄畫像及識別畫像。For example, in the example of FIG. 19 , the distance between the subject and the illumination 102 , ie, the illumination distance, is set as the photographing target condition, and the light quantity of the illumination 102 , ie, the light quantity set value, is set as the photographing device condition. In this case, the collective condition generation processing unit 313 collectively generates all combinations of the acquired values of the illumination distance and the light quantity setting value. Then, the image acquisition processing unit 212 automatically captures the images under the combined conditions of the above, and acquires the registration image and the identification image.
最適條件提示處理部314是可實行最適條件提示處理。最適條件提示處理是包含：以在總括條件產生處理產生的全部的條件來攝影的登錄畫像及識別畫像之中，針對攝取了相同的被照體的登錄畫像及識別畫像，使對照處理實行於對照處理部262，以藉由該對照處理而取得的對應點數多的攝影對象條件與攝影機器條件的組合作為最適條件提示給使用者或自動設定的處理。此情況，最適條件提示處理部314是亦可將最適條件顯示於顯示裝置12。藉此，使用者可簡單地找到攝影機器條件與攝影對象條件的最適的組合而設定。The optimum condition presentation processing unit 314 is capable of executing optimum condition presentation processing. The optimum condition presentation processing includes performing a comparison process on the registration image and the recognition image that have captured the same subject among the registration images and recognition images captured under all the conditions generated by the collective condition generation processing. The processing unit 262 is a process for presenting to the user or automatically setting the combination of the photographing target condition and the photographing device condition with a large number of corresponding points obtained by the matching process as the optimum condition. In this case, the optimum condition presentation processing unit 314 may display the optimum condition on the display device 12 . Thereby, the user can easily find and set the optimum combination of the conditions of the photographing device and the conditions of the photographic object.
例如登錄畫像及識別畫像為以圖19所示的條件來攝取相同的被照體者時，最適條件提示處理部314是從各個的畫像的對應點數來產生例如圖20所示的曲線圖，此情況，在橫軸取光量設定值亦即Lux，在縱軸取對應點數的曲線圖。而且，最適條件提示處理部314是由此曲線圖來尋找登錄畫像與識別畫像的對應點數多，且健全性(robust)高的條件的組合。此情況，可知照明距離1的A1的曲線圖的頂點的光量設定值為對應點數多且健全性最高。因此，此情況，最適條件提示處理部314是提示照明距離1與照明距離1的A1的曲線圖的頂點的光量設定值作為攝影對象條件與攝影機器條件的組合的最適條件。如此一來，最適條件提示處理部314是將攝影對象條件與攝影機器條件的最適的組合提示給使用者。另外，最適條件提示處理部314是亦可選擇幾個對應點數多且健全性高的條件的候補，推薦該等的條件作為條件提示給使用者。又，最適條件提示處理部314是亦可進行使圖20所示的曲線圖顯示於顯示裝置12的處理。For example, when the registration portrait and the identification portrait are taken under the conditions shown in FIG. 19 to capture the same subject, the optimum condition presentation processing unit 314 generates, for example, a graph as shown in FIG. 20 from the corresponding points of the respective portraits. In this case, the light quantity setting value, that is, Lux, is taken on the horizontal axis, and the graph corresponding to the number of points is taken on the vertical axis. Then, the optimum condition presentation processing unit 314 searches for a combination of conditions with a large number of corresponding points between the registration image and the recognition image from this graph and with a high robustness. In this case, it can be seen that the set value of the light quantity at the vertex of the graph of A1 of the illumination distance 1 has a large number of corresponding points and has the highest soundness. Therefore, in this case, the optimum condition presentation processing unit 314 presents the light intensity setting value at the vertex of the graph of the illumination distance 1 and A1 of the illumination distance 1 as the optimum condition for the combination of the photographing target condition and the photographing device condition. In this way, the optimum condition presentation processing unit 314 presents to the user the optimum combination of the photographing target condition and the photographing device condition. In addition, the optimum condition presentation processing unit 314 may select several candidates for conditions with a large number of corresponding points and high soundness, and recommend these conditions to be presented to the user as conditions. In addition, the optimum condition presentation processing unit 314 may perform processing for displaying the graph shown in FIG. 20 on the display device 12 .
最適條件提示處理是包含：預先準備複數個關於特徵點抽出處理的條件即特徵抽出條件，針對以在總括條件產生處理產生的全部的條件來相同的攝影後的登錄畫像及識別畫像，利用以預先準備的複數的特徵抽出條件來實行特徵點抽出處理而取得的特徵點來實行對照處理而取得的對應點數成為最多的條件的組合提示之處理。藉此，使用者可簡單地找到包含特徵抽出條件，攝影機器條件與攝影對象條件的最適的組合。The optimum condition presentation processing includes preparing a plurality of conditions for the feature point extraction processing, that is, feature extraction conditions, and using the pre-imaged registration images and identification images under all the conditions generated in the collective condition generation processing to be the same. The prepared plural feature extraction conditions are combined with the feature points obtained by executing the feature point extraction process and the corresponding conditions obtained by the matching process are presented. Thereby, the user can easily find the optimum combination including the feature extraction condition, the photographing device condition and the photographing object condition.
對照識別條件調整裝置32是具有進行在畫像對照識別裝置26的檢索處理部261及對照處理部262使用的對照識別算法的參數的調諧之機能。對照識別條件調整裝置32是具有：接受來自畫像對照識別裝置26的反餽而自動或接受來自使用者的操作而手動調整在特徵抽出裝置22或畫像對照識別裝置26的處理使用的條件或參數等，將該條件或參數指示給特徵抽出裝置22或畫像對照識別裝置26之機能。本實施形態的情況，對照識別條件調整裝置32是利用被登錄於記錄裝置24的評價用畫像登錄部242的評價用畫像，在畫像對照識別裝置26的對照處理部262進行對照，按照該對照結果此情況是對應點數來調整條件或參數。The collation recognition condition adjustment device 32 has a function of tuning the parameters of the collation recognition algorithm used by the retrieval processing unit 261 and the comparison processing unit 262 of the image comparison recognition device 26 . The comparison recognition condition adjustment device 32 is provided with: receiving feedback from the image comparison recognition device 26 to automatically adjust the conditions or parameters used in the processing of the feature extraction device 22 or the image comparison recognition device 26, or manually by accepting an operation from the user, This condition or parameter is indicated to the function of the feature extraction device 22 or the image comparison identification device 26 . In the case of the present embodiment, the comparison recognition condition adjustment device 32 uses the evaluation image registered in the evaluation image registration unit 242 of the recording device 24 to perform comparison in the comparison processing unit 262 of the image comparison recognition device 26, and according to the comparison result In this case, conditions or parameters are adjusted according to the number of points.
使用者是例如藉由操作輸入裝置11而實行對照識別條件調整裝置32的處理，可調整在特徵抽出裝置22或畫像對照識別裝置26使用的條件或參數。對照識別條件調整裝置32是如圖9所示般，具有特徵抽出條件調整處理部321、廣域特徵量字典設定處理部322、廣域特徵量相關係數計算處理部323、參數組產生處理部324及參數組設定處理部325。The user can adjust the conditions or parameters used in the feature extraction device 22 or the image comparison and identification device 26 by, for example, operating the input device 11 to execute the processing of the matching identification condition adjustment device 32 . As shown in FIG. 9 , the matching identification condition adjustment device 32 includes a feature extraction condition adjustment processing unit 321 , a wide-area feature quantity dictionary setting processing unit 322 , a wide-area feature quantity correlation coefficient calculation processing unit 323 , and a parameter group generation processing unit 324 . and the parameter group setting processing unit 325 .
特徵抽出條件調整處理部321是可實行特徵抽出條件調整處理。特徵抽出條件調整處理是包含：調整在特徵點抽出處理部221的特徵點抽出處理或在局部特徵量計算處理部222的局部特徵量計算處理所使用的特徵抽出條件之處理。使用者是可操作輸入裝置11，調整設定抽出特徵點且計算局部特徵量時的用以決定金字塔畫像的階數及角落或似曲線的臨界值等的特徵抽出條件。The feature extraction condition adjustment processing unit 321 is capable of executing feature extraction condition adjustment processing. The feature extraction condition adjustment process includes the process of adjusting the feature extraction conditions used for the feature point extraction process in the feature point extraction process unit 221 or the local feature amount calculation process in the local feature amount calculation process unit 222 . The user can operate the input device 11 to adjust and set the feature extraction conditions for determining the order of the pyramid image and the critical value of the corner or curve-like value when extracting feature points and calculating local feature quantities.
特徵抽出條件調整處理是包含：根據畫像對照識別裝置26的對照處理部262的對照處理的結果，針對局部特徵量計算處理所致的特徵抽出條件，調整為正對應點數與負對應點數的比會形成最大的處理。在本實施形態中，所謂正對應點是如圖18(A)所示般，意思被登錄於記錄裝置24的評價用畫像登錄部242的登錄畫像群之中，攝取了與識別畫像的被照體相同的被照體的登錄畫像與識別畫像的對應點。而且，所謂正對應點數是意思正對應點的數量。圖18(A)的例子是在登錄畫像所拍攝的被照體W1與在識別畫像所拍攝的被照體W1是相同。The feature extraction condition adjustment process includes adjusting the feature extraction condition by the local feature amount calculation process to the number of positive correspondence points and the number of negative correspondence points, based on the result of the comparison process performed by the comparison processing unit 262 of the image comparison recognition device 26. ratio will form the largest processing. In the present embodiment, the “positive correspondence point” means that as shown in FIG. 18(A) , it is registered in the registered image group of the evaluation image registration unit 242 of the recording device 24, and the photographed image corresponding to the identification image is captured. The corresponding points of the registered portrait and the identification portrait of the subject with the same body. In addition, the number of directly corresponding points means the number of directly corresponding points. In the example of FIG. 18(A) , the subject W1 captured in the registration image and the subject W1 captured in the recognition image are the same.
又，所謂負對應點是如圖18(B)所示般，意思被登錄於記錄裝置24的評價用畫像登錄部242的登錄畫像群之中，攝取了與識別畫像的被照體不同的被照體的登錄畫像與識別畫像的對應點。而且，所謂負對應點數是意思負對應點的數量。圖18(B)的例子是在登錄畫像所拍攝的被照體W2與在識別畫像所拍攝的被照體W1是相異者。In addition, the so-called negative corresponding point means that, as shown in FIG. 18(B) , it is registered in the registered image group of the evaluation image registration unit 242 of the recording device 24, and a subject different from the subject of the recognition image is captured. The corresponding point of the registered portrait of the photograph and the identification portrait. In addition, the number of negative corresponding points means the number of negative corresponding points. In the example of FIG. 18(B) , the subject W2 captured in the registration image and the subject W1 captured in the recognition image are different.
廣域特徵量字典設定處理部322是可實行廣域特徵量字典設定處理。廣域特徵量字典設定處理是包含：按每個廣域特徵量計算統計資訊來計算得分(score)，自動設定其得分高者作為用在局部特徵量群分類處理部223的局部特徵量群分類處理的廣域特徵量字典之處理。所謂廣域特徵量計算統計資訊是在藉由廣域特徵量計算處理來計算廣域特徵量的過程，亦即在局部特徵量群分類處理部223的局部特徵量群分類處理的過程所取得的統計資訊，能以投票給與局部特徵量相關性高的字典資訊而得到的柱狀圖或相關值的統計量的得分來表示。可思考根據特定的字典資訊被選的次數等的統計資訊之得分高的字典資訊是性能高。若根據廣域特徵量字典設定處理，則可自動選擇具有如此的性能高的字典資訊之廣域特徵量字典。The wide-area feature amount dictionary setting processing unit 322 is capable of executing the wide-area feature amount dictionary setting process. The wide-area feature amount dictionary setting process includes: calculating statistical information for each wide-area feature amount to calculate a score, and automatically setting a higher score as the local feature amount group classification used in the local feature amount group classification processing unit 223 The processing of the processed wide-area feature dictionary. The so-called wide-area feature amount calculation statistical information is obtained in the process of calculating the wide-area feature amount by the wide-area feature amount calculation process, that is, in the process of the local feature amount group classification process in the local feature amount group classification processing unit 223. The statistical information can be represented by a histogram obtained by voting for dictionary information having a high correlation with the local feature amount, or a score of a statistical amount of the correlation value. It can be considered that dictionary information with a high score based on statistical information such as the number of times specific dictionary information is selected has high performance. If the process is set according to the wide-area feature dictionary, a wide-area feature dictionary having such high-performance dictionary information can be automatically selected.
廣域特徵量字典設定處理是包含：選擇被登錄於廣域特徵量字典登錄部245的複數的廣域特徵量字典之中，根據廣域特徵量相關係數，可將攝取了相同的被照體的登錄畫像與識別畫像判斷成相同，將攝取了不同的被照體的登錄畫像與識別畫像判斷成不相同的廣域特徵量字典，設定成用在局部特徵量群分類處理的廣域特徵量字典之處理。The wide-area feature value dictionary setting process includes selecting a plurality of wide-area feature value dictionaries registered in the wide-area feature value dictionary registration unit 245, and based on the wide-area feature value correlation coefficient, the same subject can be captured. The registration image and the recognition image are determined to be the same, and the wide-area feature quantity dictionary in which the registration image and the recognition image of different subjects are judged to be different is set as the wide-area feature quantity used in the classification processing of the local feature quantity group. Handling of dictionaries.
廣域特徵量相關係數是成為表示登錄畫像與識別畫像的廣域特徵量的一致度亦即性能之一的指標。意思2個的畫像間的廣域特徵量的相關性越大，在兩者的畫像含有相同的被照體的可能性越高。因此，攝取了相同的被照體的畫像間的廣域特徵量的相關性大，是意思針對攝取了該相同的被照體的畫像，使用其廣域特徵量字典來計算的廣域特徵量的一致度高。亦即，這是意思在攝取了相同的被照體的畫像間其廣域特徵量字典的性能高。The correlation coefficient of the wide-area feature amount is an index that represents the degree of agreement between the wide-area feature amounts of the registration profile and the recognition profile, that is, one of the performances. This means that the greater the correlation between the wide-area feature quantities between the two images, the higher the possibility that the two images contain the same subject. Therefore, the correlation between the wide-area feature quantities between the images that capture the same subject is high, which means that the wide-area feature quantities are calculated using the wide-area feature quantity dictionary for the images that capture the same subject. of high consistency. That is, this means that the performance of the wide-area feature-quantity dictionary is high among the images that have captured the same subject.
另一方面，意思2個的畫像間的廣域特徵量的相關性越小，在兩者的畫像中所含的被照體不同的可能性越高。因此，攝取了不同的被照體的畫像間的廣域特徵量的相關性小，是意思針對攝取了該不同的被照體的畫像，使用其廣域特徵量字典來計算的廣域特徵量的一致度低。亦即，這是意思在攝取了不同的被照體的畫像間其廣域特徵量字典的性能高。On the other hand, the smaller the correlation between the wide-area feature amounts between the two images, the higher the possibility that the subjects included in the two images are different. Therefore, the correlation between the wide-area feature quantities between the images that have captured different subjects is small, which means that the wide-area feature quantities are calculated using the wide-area feature quantity dictionary for the images that have captured the different subjects. The consistency is low. That is, this means that the performance of the wide-area feature-quantity dictionary is high between images of different subjects.
廣域特徵量字典設定處理是例如其次般計算廣域特徵量相關係數，可根據該廣域特徵量相關係數來判斷廣域特徵量字典的性能。亦即，廣域特徵量字典設定處理是當2個的廣域特徵量為向量V時，將各個的絶對值正規化成1，取內積的值，針對該內積的值接近1者判斷成同一性高，針對該內積的值接近-1者判斷成同一性低。In the wide-area feature amount dictionary setting process, for example, the wide-area feature amount correlation coefficient is calculated next, and the performance of the wide-area feature amount dictionary can be judged based on the wide-area feature amount correlation coefficient. That is, in the wide-area feature amount dictionary setting process, when two wide-area feature amounts are vectors V, the absolute value of each is normalized to 1, the value of the inner product is taken, and if the value of the inner product is close to 1, it is determined as The identity is high, and if the value of the inner product is close to -1, the identity is judged to be low.
此情況，例如將2個的廣域特徵量之中1個的廣域特徵量向量Va設為其次的(2)式，將被正規化的廣域特徵量向量Vb設為其次的(3)式，將d定義為自然數。此情況，被正規化的廣域特徵量向量Vb是利用廣域特徵量向量Va來以其次的(4)式表示。然後，若將廣域特徵量向量Va設為其次的(5)式，則廣域特徵量向量Va的絶對值是成為其次的(6)式。 In this case, for example, the wide-area feature vector Va of one of the two wide-area feature values is the second formula (2), and the normalized wide-area feature vector Vb is the second formula (3) formula, d is defined as a natural number. In this case, the normalized wide-area feature vector Vb is expressed by the following equation (4) using the wide-area feature vector Va. Then, if the wide-area feature vector Va is the next formula (5), the absolute value of the wide-area feature vector Va is the next formula (6).
然後，若將2個的廣域特徵量之中另一個的廣域特徵量向量Wa設為其次的(7)式，則其內積的值是成為其次的(8)式。然後，廣域特徵量字典登錄部245針對以(8)式所取得的內積的值亦即廣域特徵量相關係數接近1者是判斷成同一性高亦即相同的被照體的可能性高，針對該內積的值接近-1者則是判斷成同一性低亦即相同的被照體的可能性低。 Then, if the other wide-area feature vector Wa of the two wide-area feature amounts is the second formula (7), the value of the inner product becomes the second formula (8). Then, the wide-area feature value dictionary registration unit 245 determines that the value of the inner product obtained by the formula (8), that is, the value of the wide-area feature value correlation coefficient is close to 1, and it is determined that the identity is high, that is, the same subject is likely to be photographed. If the value of the inner product is close to -1, it is judged that the identity is low, that is, the possibility of the same subject is low.
又，廣域特徵量字典設定處理是例如其次般計算廣域特徵量相關係數，可根據該廣域特徵量相關係數來判斷廣域特徵量字典的性能。亦即，廣域特徵量相關係數是設為2個的廣域特徵量的向量間的距離的值。然後，廣域特徵量字典設定處理是若2個的廣域特徵量的向量間的距離的值即廣域特徵量相關係數接近0，則判斷成同一性高，值大時則判斷成同一性低。In addition, in the wide-area feature amount dictionary setting process, for example, the wide-area feature amount correlation coefficient is calculated next, and the performance of the wide-area feature amount dictionary can be judged based on the wide-area feature amount correlation coefficient. That is, the wide-area feature quantity correlation coefficient is a value of the distance between two broad-area feature quantity vectors. Then, in the wide-area feature amount dictionary setting process, if the value of the distance between the vectors of the two wide-area feature amounts, that is, the wide-area feature amount correlation coefficient is close to 0, the identity is determined to be high, and if the value is large, the identity is determined to be high. Low.
此情況，例如將2個的其中一方的廣域特徵量向量Va設為上述(2)、(5)式，且將另一方的廣域特徵量向量Wa設為上述(7)及下述(9)式時，廣域特徵量向量Va、Wa間的距離是以其次的(10)式來表示。 In this case, for example, one of the two broad-area feature vector Va is set to the above-mentioned equations (2) and (5), and the other wide-area feature vector Wa is set to the above-mentioned (7) and the following ( In the formula 9), the distance between the wide-area feature vector Va and Wa is represented by the following formula (10).
然後，廣域特徵量字典登錄部245針對以(10)式所取得的向量Va、Wa間的距離近亦即接近0者是判斷成同一性高亦即相同的被照體的能性高，針對向量Va、Wa間的距離大者則是判斷成同一性低亦即相同的被照體的可能性低。另外，此時各個的向量Va、Wa是不一定要被正規化。Then, the wide-area feature value dictionary registration unit 245 judges that the distance between the vectors Va and Wa obtained by the formula (10) is close, that is, close to 0, that the identity is high, that is, the possibility of the same subject is high, If the distance between the vectors Va and Wa is large, it is judged that the identity is low, that is, the possibility of the same subject is low. In addition, at this time, the respective vectors Va and Wa do not necessarily need to be normalized.
廣域特徵量相關係數計算處理部323是可實行廣域特徵量相關係數計算處理。廣域特徵量相關係數計算處理是包含：比較各登錄畫像的廣域特徵量與識別畫像的廣域特徵量，計算表示其相關關係的廣域特徵量相關係數之處理。例如，廣域特徵量相關係數計算處理是可比較各登錄畫像的廣域特徵量與識別畫像的廣域特徵量的柱狀圖(histogram)來計算廣域特徵量相關係數。The wide-area feature amount correlation coefficient calculation processing unit 323 is capable of executing a wide-area feature amount correlation coefficient calculation process. The wide-area feature amount correlation coefficient calculation process includes comparing the wide-area feature amount of each registered profile with the wide-area feature amount of the recognition profile, and calculating a wide-area feature amount correlation coefficient indicating the correlation. For example, in the wide-area feature value correlation coefficient calculation process, the wide-area feature value correlation coefficient can be calculated by comparing the wide-area feature value of each registered image with a histogram of the wide-area feature value of the recognition image.
參數組產生處理部324是可實行參數組產生處理。參數組產生處理是包含：例如圖17所示般，針對用以設定在畫像對照識別裝置26的對照處理部262所實行的對照處理的條件之參數，組合被設定成各參數所定的設定範圍內的設定值之各參數來產生複數的參數組之處理。本實施形態的情況，參數組產生處理是產生有關各參數的設定值的全部的組合的參數組。The parameter group generation processing unit 324 is capable of executing parameter group generation processing. The parameter group generation process includes, for example, as shown in FIG. 17 , a combination of parameters for setting the conditions of the comparison processing performed by the comparison processing unit 262 of the image comparison recognition device 26 is set within the setting range specified by each parameter. The processing of each parameter of the set value to generate a complex parameter group. In the case of the present embodiment, the parameter group generation process is to generate a parameter group of all combinations of the set values of each parameter.
參數組設定處理部325是可實行參數組設定處理。參數組設定處理是包含：根據使用者的操作，設定使用在檢索處理部261的檢索處理的參數組之處理。使用者是例如可操作輸入裝置11，設定使用在檢索處理部261的檢索處理的參數組。The parameter group setting processing unit 325 is capable of executing the parameter group setting processing. The parameter set setting process includes a process of setting a parameter set used in the search process of the search processing unit 261 according to the operation of the user. For example, the user can operate the input device 11 to set a parameter group used in the retrieval processing of the retrieval processing unit 261 .
可視化裝置33是具有將從畫像對照識別裝置26接受的資訊、或關於攝影條件調整裝置31及對照識別條件調整裝置32的調整結果的資訊等的各種的資訊可視化而顯示於顯示裝置12的機能。在本實施形態中所謂可視化是不僅將資料形成文字或數值，還包含形成圖形或圖表等而顯示於顯示裝置12的情形。可視化裝置33是如圖10所示般，具有特徵點顯示處理部331、廣域特徵量計算統計資訊顯示處理部332、對照結果顯示處理部333及條件顯示處理部334。The visualization device 33 has a function of visualizing and displaying on the display device 12 various kinds of information, such as information received from the image matching and identifying device 26 , or information about the adjustment results of the photographing condition adjusting device 31 and the matching identifying condition adjusting device 32 . In the present embodiment, the term "visualization" includes not only characters or numerical values of data, but also graphics, graphs, and the like, which are displayed on the display device 12 . As shown in FIG. 10 , the visualization device 33 includes a feature point display processing unit 331 , a wide-area feature amount calculation statistical information display processing unit 332 , a comparison result display processing unit 333 , and a condition display processing unit 334 .
特徵點顯示處理部331是可實行特徵點顯示處理。特徵點顯示處理是可在藉由攝影條件調整裝置31或對照識別條件調整裝置32來調整各種的條件或參數時實行。特徵點顯示處理是包含：例如圖21(A)所示般，將藉由特徵抽出裝置22的特徵點抽出處理部221的特徵點抽出處理的實行所抽出的各特徵點P重疊於畫像而顯示於顯示裝置12的處理。亦即，特徵點P會被重疊顯示於畫像中的被照體S上。藉此，使用者是可目視確認是否被抽出畫像中的被照體S的何處作為特徵點。又，特徵點顯示處理是包含：如圖21(A)所示般，以按局部特徵量所屬的每個局部特徵量群，此情況，按各特徵點P群集充當的每個字典資訊d而異的形態，例如不同的顏色或大小、形狀來將各特徵點P顯示於顯示裝置12的處理。在圖21中，以濃淡的不同來表示各特徵點P的色彩的不同。The feature point display processing unit 331 is capable of executing feature point display processing. The feature point display process can be executed when various conditions or parameters are adjusted by the imaging condition adjustment means 31 or the comparison recognition condition adjustment means 32 . The feature point display process includes, for example, as shown in FIG. 21(A) , superimposing and displaying each feature point P extracted by the execution of the feature point extraction process of the feature point extraction processing unit 221 of the feature extraction device 22 on the image. processing on the display device 12 . That is, the feature point P will be superimposed and displayed on the subject S in the portrait. Thereby, the user can visually confirm whether or not the subject S in the image is extracted as a feature point. Also, the feature point display process includes: as shown in FIG. 21(A), for each local feature value group to which the local feature value belongs, in this case, for each dictionary information d serving as each feature point P cluster A process of displaying each feature point P on the display device 12 in a different form, such as a different color, size, or shape. In FIG. 21 , the difference in color of each feature point P is represented by the difference in gradation.
廣域特徵量計算統計資訊顯示處理部332是可實行廣域特徵量計算統計資訊顯示處理。廣域特徵量計算統計資訊顯示處理是可在藉由攝影條件調整裝置31或對照識別條件調整裝置32來調整各種的條件或參數時實行。廣域特徵量計算統計資訊顯示處理是包含：將廣域特徵量計算統計資訊顯示於顯示裝置12的處理。廣域特徵量計算統計資訊是藉由在特徵抽出裝置22的廣域特徵量計算處理部224所實行的廣域特徵量計算處理來計算廣域特徵量的過程取得的統計資訊。廣域特徵量計算統計資訊顯示處理是包含：如圖21(B)所示般，以投票給與各局部特徵量A相關性高的字典資訊d而得到的柱狀圖或相關值的統計量的得分作為廣域特徵量計算統計資訊G來顯示於顯示裝置12的處理。亦即，廣域特徵量計算統計資訊G是可設為在廣域特徵量計算處理的實行過程取得的廣域特徵量的柱狀圖或相關值的統計量的得分。The wide-area feature amount calculation statistical information display processing unit 332 is capable of executing wide-area feature amount calculation statistical information display processing. The wide-area feature amount calculation statistical information display processing can be performed when various conditions or parameters are adjusted by the imaging condition adjustment means 31 or the comparison recognition condition adjustment means 32 . The wide-area feature amount calculation statistical information display processing includes a process of displaying the wide-area feature amount calculation statistical information on the display device 12 . The wide-area feature amount calculation statistical information is statistical information obtained by the process of calculating the wide-area feature amount by the wide-area feature amount calculation process performed by the wide-area feature amount calculation processing unit 224 of the feature extraction device 22 . The wide-area feature amount calculation statistical information display process includes: as shown in FIG. 21(B) , a histogram or correlation value statistics obtained by voting for dictionary information d that has a high correlation with each local feature amount A The score of , is a process of displaying on the display device 12 as the wide-area feature amount calculation statistic information G. That is, the wide-area feature amount calculation statistic information G is a score that can be a histogram of the wide-area feature amount or a statistic of correlation values acquired during the execution of the wide-area feature amount calculation process.
對照結果顯示處理部333是可實行對照結果顯示處理。對照結果顯示處理是可在藉由攝影條件調整裝置31或對照識別條件調整裝置32來調整各種的條件或參數時實行。對照結果顯示處理是包含：將在畫像對照識別裝置26的對照處理部262所實行的複數的對照處理的結果彙整而可視化顯示於顯示裝置12之處理。The collation result display processing unit 333 is capable of executing collation result display processing. The comparison result display process can be executed when various conditions or parameters are adjusted by the photographing condition adjustment means 31 or the comparison recognition condition adjustment means 32 . The collation result display process includes a process of integrating the results of the plural collation processes performed by the collation processing unit 262 of the image collation recognition device 26 and displaying them on the display device 12 as a visualization.
對照結果顯示處理是包含：例如將利用圖17所示的各參數組來實行對照處理的對照結果如圖22所示般圖表化而顯示的處理。此情況，對照結果顯示處理部333是不僅相同的被照體的畫像彼此間，有關不同的被照體的畫像彼此間也是進行對照處理。然後，對照結果顯示處理部333是分別計算：攝取了相同的被照體的畫像間的對應點數即正對應點數，及攝取了不同的被照體的畫像間的對應點數即負對應點數，而圖表化顯示於顯示裝置12。圖22的圖表是在橫軸取條件No，在縱軸取對應點數者。又，圖22之中，F1是表示正對應點數的圖表，F2是表示負對應點數的圖表。就圖22的例子而言，可知條件No.N是正對應點數最多且負對應點數最少。藉此，使用者可一看就能掌握正對應點數多且負對應點數少的條件No亦即性能佳的條件No。The comparison result display process includes, for example, a process of displaying the comparison result obtained by performing the comparison process using each parameter group shown in FIG. 17 in a graph as shown in FIG. 22 . In this case, the comparison result display processing unit 333 performs comparison processing not only between the images of the same subject, but also between the images of different subjects. Then, the comparison result display processing unit 333 separately calculates the number of correspondence points between the images of the same subject, that is, the number of positive correspondences, and the number of correspondence points between the images of the images of different subjects, that is, the number of negative correspondences. The number of points is displayed on the display device 12 as a graph. In the graph of FIG. 22, the condition No. is taken on the horizontal axis, and the corresponding points are taken on the vertical axis. In addition, in FIG. 22, F1 is a graph showing the number of positive correspondence points, and F2 is a graph showing the number of negative correspondence points. In the example of FIG. 22 , it can be seen that the condition No. N has the largest number of positive correspondence points and the smallest number of negative correspondence points. As a result, the user can grasp the condition No of a large number of positive correspondence points and a small number of negative correspondence points, that is, a condition No of good performance at a glance.
又，對照結果顯示處理是包含：在畫像對照識別裝置26的對照處理部262中實行對照處理時，將用在對照的識別畫像的對應點數評價資訊顯示於顯示裝置12的處理。對應點數評價資訊是根據在對照處理計算的對應點數之資訊，例如可為對應點數本身的數值，或亦可例如設為複數階段的等級或得分。The matching result display processing includes processing of displaying on the display device 12 the evaluation information corresponding to the number of points of the identified image used for matching when the matching processing unit 262 of the image matching recognition device 26 executes the matching processing. The corresponding point evaluation information is information based on the corresponding points calculated in the comparison process, and may be, for example, the value of the corresponding points itself, or may be, for example, a level or score of a plurality of stages.
此情況，對照結果顯示處理部333是例如圖23所示般，在顯示裝置12的畫面上，使登錄畫像顯示部51、真值畫像顯示區域52、偽值畫像顯示區域53、真值畫像評價資訊顯示區域541、真值畫像相關資訊顯示區域542、偽值畫像評價資訊顯示區域551及偽值畫像相關資訊顯示區域552顯示。在登錄畫像顯示部51是顯示進行對照處理的畫像之中，被登錄於記錄裝置24的評價用畫像登錄部242的登錄畫像群之中的1個 。In this case, the comparison result display processing unit 333 displays, on the screen of the display device 12, the registration image display unit 51, the true value image display area 52, the false value image display area 53, and the true value image evaluation as shown in FIG. 23, for example. The information display area 541, the information display area 542 related to the true value portrait, the evaluation information display area 551 of the pseudo value portrait, and the information display area 552 related to the pseudo value portrait are displayed. The registered portrait display unit 51 displays one of the registered portraits registered in the evaluation portrait registration unit 242 of the recording device 24 among the portraits subjected to the matching process.
在真值畫像顯示區域52是顯示進行對照處理的畫像之中，被登錄於記錄裝置24的評價用畫像登錄部242的識別畫像，以不同的條件攝取了與被顯示於登錄畫像顯示部51的畫像的被照體相同的被照體的畫像。就圖23的例子而言，被顯示於登錄畫像顯示部51的畫像的被照體W1與被顯示於真值畫像顯示區域52的畫像的被照體W1是相同的個體。在本實施形態中，將以不同的條件攝取了與被顯示於登錄畫像顯示部51的畫像的被照體相同的被照體的畫像稱為真值畫像。此情況，攝取了相同的被照體的登錄畫像與識別畫像是在相互建立關聯的狀態亦即組成一對的狀態下被登錄於評價用畫像登錄部242。亦即，對照結果顯示處理部333是藉由識別登錄畫像與識別畫像的的關聯，可認知攝取了相同的被照體者或攝取了不同的被照體者。In the true value image display area 52, among the images for which the comparison processing is performed, the identification images registered in the evaluation image registration unit 242 of the recording device 24 are captured under different conditions from those displayed in the registration image display unit 51. A portrait of the same subject as the subject of the portrait. In the example of FIG. 23 , the subject W1 of the image displayed on the registered image display unit 51 and the subject W1 of the image displayed in the true value image display area 52 are the same individual. In the present embodiment, an image in which the same subject as the image displayed on the registered image display unit 51 is captured under different conditions is referred to as a true image. In this case, the registration image and the identification image that have captured the same subject are registered in the evaluation image registration unit 242 in a state in which they are associated with each other, that is, in a paired state. That is, the comparison result display processing unit 333 recognizes that the same subject or a different subject is captured by recognizing the association between the registered portrait and the identification portrait.
而且，在偽值畫像顯示區域53是顯示進行對照處理的畫像之中，被登錄於記錄裝置24的評價用畫像登錄部242的識別畫像，以不同的條件攝取了與被顯示於登錄畫像顯示部51的畫像的被照體不同的被照體的畫像。就圖23的例子而言，被顯示於登錄畫像顯示部51的畫像的被照體W1與被顯示於真值畫像顯示區域52的畫像的被照體W2是不同的個體。在本實施形態中，將攝取了與被顯示於登錄畫像顯示部51的畫像的被照體不同的被照體的畫像稱為偽值畫像。Further, in the pseudo-value image display area 53, among the images for which the collation process is displayed, the identification images registered in the evaluation image registration unit 242 of the recording device 24 are captured under different conditions from those displayed in the registration image display unit. 51 The portrait of the subject is a portrait of a different subject. In the example of FIG. 23 , the subject W1 of the image displayed on the registered image display unit 51 and the subject W2 of the image displayed in the true value image display area 52 are different individuals. In the present embodiment, an image in which a subject different from the subject in the image displayed on the registered image display unit 51 is captured is referred to as a dummy image.
又，在真值畫像評價資訊顯示區域541是顯示以被顯示於登錄畫像顯示部51的畫像及被顯示於真值畫像顯示區域52的畫像來進行對照處理的結果，亦即真值畫像的對應點數評價資訊。又，在偽值畫像評價資訊顯示區域551是顯示以被顯示於登錄畫像顯示部51的畫像及被顯示於偽值畫像顯示區域53的畫像來進行對照處理的結果，亦即偽值畫像的對應點數評價資訊。In addition, in the real-value image evaluation information display area 541, the result of the comparison processing with the image displayed on the registration image display unit 51 and the image displayed in the real-value image display area 52 is displayed, that is, the correspondence of the real-value image is displayed. Point evaluation information. In addition, in the pseudo-value image evaluation information display area 551, the result of the comparison process with the image displayed on the registration image display unit 51 and the image displayed in the pseudo-value image display area 53 is displayed, that is, the correspondence of the pseudo-value image is displayed. Point evaluation information.
對照結果顯示處理是包含：根據廣域特徵量相關係數，將資訊顯示於顯示裝置12的處理。此情況，在真值畫像相關資訊顯示區域542是顯示由被顯示於登錄畫像顯示部51的登錄畫像來計算的廣域特徵量與由被顯示於真值畫像顯示區域52的真值畫像來計算的廣域特徵量的相關係數。而且，在偽值畫像相關資訊顯示區域552是顯示比較由被顯示於登錄畫像顯示部51的登錄畫像來計算的廣域特徵量與由被顯示於偽值畫像顯示區域53的偽值畫像來計算的廣域特徵量而得到的相關係數。The collation result display process includes a process of displaying information on the display device 12 based on the correlation coefficient of the wide-area feature amount. In this case, in the information display area 542 related to the true value image, the wide-area feature amount calculated from the registered image displayed on the registered image display unit 51 and the value calculated from the true value image displayed in the true value image display area 52 are displayed. The correlation coefficient of the wide-area feature quantity. In addition, in the pseudo-value image-related information display area 552, a comparison between the wide-area feature amount calculated from the registration image displayed on the registration image display unit 51 and the pseudo-value image displayed in the pseudo-value image display area 53 is displayed. The correlation coefficient obtained by the wide-area feature quantity of .
使用者是例如操作輸入裝置11，選擇顯示於登錄畫像顯示部51、真值畫像顯示區域52、偽值畫像顯示區域53、真值畫像評價資訊顯示區域541的畫像，亦即進行對照處理的畫像，針對該畫像實行對照處理。然後，一旦對照處理被進行，則對照結果顯示處理部333是分別將對應點數評價資訊作為其對照結果顯示於真值畫像評價資訊顯示區域541及偽值畫像評價資訊顯示區域551，且分別將廣域特徵量相關係數顯示於真值畫像相關資訊顯示區域542及偽值畫像相關資訊顯示區域552。藉此，使用者藉由看根據被顯示於真值畫像評價資訊顯示區域541及偽值畫像評價資訊顯示區域551的對應點數評價資訊、或根據被顯示於真值畫像相關資訊顯示區域542及偽值畫像相關資訊顯示區域552的廣域特徵量相關係數的資訊，可確認用在對照處理的各種條件或參數的性能。此情況，所謂根據廣域特徵量相關係數的資訊是亦可為廣域特徵量相關係數本身，或亦可為將廣域特徵量相關係數加工而以數階段表示性能之類的等級或排序等。For example, the user operates the input device 11 and selects the images displayed on the login image display unit 51, the true value image display area 52, the false value image display area 53, and the true value image evaluation information display area 541, that is, the image for which the comparison processing is performed. , and carry out the comparison processing for the portrait. Then, once the collation process is performed, the collation result display processing unit 333 displays the corresponding point evaluation information as the collation result in the real-value image evaluation information display area 541 and the pseudo-value image evaluation information display area 551, respectively, and The correlation coefficients of the wide-area feature quantities are displayed in the real-valued image-related information display area 542 and the pseudo-valued image-related information display area 552 . Thereby, the user can evaluate the information according to the corresponding points displayed in the real-value portrait evaluation information display area 541 and the pseudo-value portrait evaluation information display area 551, or according to the information displayed in the real-value portrait-related information display area 542 and The information on the correlation coefficient of the wide-area feature quantity in the pseudo-value portrait-related information display area 552 can confirm the performance of various conditions or parameters used in the matching process. In this case, the information based on the correlation coefficient of the wide-area feature quantity may be the correlation coefficient of the wide-area feature quantity itself, or may be a rank or order such as the performance in several stages by processing the correlation coefficient of the wide-area feature quantity. .
對照結果顯示處理部333是例如圖18所示般，亦可使登錄畫像與識別畫像的正對應點數或負對應點數顯示於顯示裝置12。又，對照結果顯示處理部333是亦可例如在登錄畫像及識別畫像，將各個的畫像所分別固有的特徵點與共有的特徵點分開顯示。此情況，對照結果顯示處理部333是亦可顯示各個的畫像所固有的特徵點與共有的特徵點的其數量或比例。而且，對照結果顯示處理部333是亦可任意地切換：只顯示登錄畫像及識別畫像的模式、例如在登錄畫像及識別畫像追加特徵點而顯示的模式、和在登錄畫像及識別畫像追加特徵點與對應點而顯示的模式。As shown in FIG. 18 , the comparison result display processing unit 333 may display on the display device 12 the positive or negative corresponding points of the registration image and the recognition image. In addition, the comparison result display processing unit 333 may, for example, separately display the feature points unique to the respective images and the common feature points in the registration image and the identification image. In this case, the comparison result display processing unit 333 may also display the number or ratio of the characteristic points unique to each image and the common characteristic points. In addition, the matching result display processing unit 333 can be arbitrarily switched between a mode in which only the registration image and the recognition image are displayed, a mode in which feature points are added to the registration image and the recognition image, and a mode in which feature points are added to the registration image and the recognition image. The mode displayed with the corresponding point.
在此，如工業製品等般，在同一的工程被製造的零件，相較於人的臉或指紋等的生物特徵認證，個體間的特徵的差極小。因此，需要對照被顯現於個體表面的多數的特徵，因此計算量多，在識別花時間，而且識別對象的母數越增加，計算量越爆發性地增加。Here, like an industrial product or the like, a part manufactured in the same process has a very small difference in characteristics between individuals compared to biometric authentication such as a human face or a fingerprint. Therefore, it is necessary to compare many features appearing on the surface of an individual, so the amount of calculation is large, and it takes time to recognize, and the more the number of identification objects is increased, the more the amount of calculation increases explosively.
對於此，個體識別系統1是具備：畫像取得處理部212、特徵點抽出處理部221、局部特徵量計算處理部222、局部特徵量群分類處理部223、檢索對象畫像登錄處理部231、廣域特徵量登錄處理部234、鎖定處理部2611及特定處理部2612。畫像取得處理部212是可實行取得藉由攝影機器10來攝影的被照體的畫像之畫像取得處理。特徵點抽出處理部221是可實行從在畫像取得處理取得的畫像抽出特徵點之特徵點抽出處理。局部特徵量計算處理部222是可實行計算在特徵點抽出處理被抽出的特徵點的局部特徵量之局部特徵量計算處理。In this regard, the individual identification system 1 includes an image acquisition processing unit 212, a feature point extraction processing unit 221, a local feature value calculation processing unit 222, a local feature value group classification processing unit 223, a search target image registration processing unit 231, a wide area The feature registration processing unit 234 , the lock processing unit 2611 , and the specific processing unit 2612 . The image acquisition processing unit 212 is an image acquisition process capable of acquiring an image of the subject photographed by the photographing device 10 . The feature point extraction processing unit 221 is a feature point extraction process capable of extracting feature points from an image acquired in the image acquisition process. The local feature amount calculation processing unit 222 is capable of executing a local feature amount calculation process for calculating the local feature amount of the feature point extracted by the feature point extraction process.
局部特徵量群分類處理部223是可實行局部特徵量群分類處理，將藉由局部特徵量計算處理所取得的複數的局部特徵量予以按照該局部特徵量的值來分類成預定數例如64個的局部特徵量群。廣域特徵量計算處理部224是可實行根據各局部特徵量群來計算廣域特徵量的廣域特徵量計算處理。檢索對象畫像登錄處理部231是可實行預先以複數的檢索對象的畫像作為登錄畫像來登錄於檢索對象畫像登錄部241的檢索對象畫像登錄處理。廣域特徵量登錄處理部234是可實行將登錄畫像的廣域特徵量登錄於廣域特徵量登錄部244的廣域特徵量登錄處理。鎖定處理部2611是可實行鎖定處理，例如圖15所示般，從複數的登錄畫像之中，以具有與識別畫像所具有的廣域特徵量的相關性高的廣域特徵量之預定數的登錄畫像作為候補鎖定。然後，特定處理部2612是可實行特定處理，例如圖16所示般，比較藉由鎖定處理所抽出的候補的登錄畫像的局部特徵量與識別畫像的局部特徵量，特定局部特徵量的對應點數最多的登錄畫像。The local feature value group classification processing unit 223 is capable of performing local feature value group classification processing, and classifies the complex number of local feature values obtained by the local feature value calculation process into a predetermined number, for example, 64 according to the value of the local feature value. The local feature group of . The wide-area feature amount calculation processing unit 224 is a wide-area feature amount calculation process capable of calculating a wide-area feature amount from each local feature amount group. The retrieval-target image registration processing unit 231 is a retrieval-target image registration process capable of registering a plurality of retrieval-target images as registration images in the retrieval-target image registration unit 241 in advance. The wide-area feature value registration processing unit 234 is capable of executing a wide-area feature value registration process for registering the wide-area feature value of the registered image in the wide area feature value registration unit 244 . The lock processing unit 2611 is capable of executing lock processing, for example, as shown in FIG. 15 , from among a plurality of registration portraits, a predetermined number of wide-area feature values having a high correlation with the wide-area feature amount possessed by the identification portrait are selected. The login image is locked as an alternate. Then, the specific processing unit 2612 can execute specific processing. For example, as shown in FIG. 16 , it compares the local feature value of the registration image of the candidate extracted by the lock processing with the local feature value of the recognition image, and specifies the corresponding point of the local feature value. The largest number of login images.
若根據此構成，則鎖定處理部2611是藉由鎖定處理的實行，比較登錄畫像與識別畫像的廣域特徵量來大致地鎖定成為候補的登錄畫像。此鎖定處理是相較於全部對比各個的畫像的局部特徵量的特定處理，計算量少非常高速處理，但難取得精度。於是其次，特定處理部2612是藉由特定處理的實行，對比被鎖定的候補的登錄畫像與識別畫像的局部特徵量來特定對象。此特定處理是相較於以廣域特徵量來比較的鎖定處理，雖計算量多花時間，但精度佳。如此，個體識別系統1是藉由以廣域特徵量來鎖定的鎖定處理及以局部特徵量來特定的特定處理的2階段的處理，可實現高速且高精度的檢索。這在登錄畫像的母數增加的情況特別發揮效果。According to this configuration, the lock processing unit 2611 roughly locks a candidate registration image by comparing the wide-area feature amount of the registration image and the recognition image by executing the lock process. This locking process is a very high-speed process with a small amount of calculation compared to a specific process that compares the local feature values of each of the images in total, but it is difficult to obtain accuracy. Then, the identification processing unit 2612 executes the identification processing to identify the object by comparing the registered portrait of the locked candidate with the local feature value of the identification portrait. This specific processing is more accurate than the locking processing compared with the wide-area feature quantity, although it takes more time to calculate. In this way, the individual identification system 1 can realize high-speed and high-accuracy retrieval by the two-stage process of the locking process for locking with the wide-area feature amount and the specific process for specifying with the local feature amount. This is particularly effective when the number of characters of the registered image is increased.
而且，個體識別系統1是更具備攝影機器條件調整處理部311、攝影對象條件調整處理部312及特徵抽出條件調整處理部321之中至少任一個。攝影機器條件調整處理部311是可實行攝影機器條件調整處理，根據使用者的操作調整有關攝影機器10的條件即攝影機器條件。攝影對象條件調整處理部312是可實行攝影對象條件調整處理，根據使用者的操作調整有關攝影對象的條件即攝影對象條件。而且，特徵抽出條件調整處理部321是可實行特徵抽出條件調整處理，根據使用者的操作調整有關特徵點抽出處理的條件即特徵抽出條件。Furthermore, the individual identification system 1 further includes at least any one of the imaging device condition adjustment processing unit 311 , the imaging target condition adjustment processing unit 312 , and the feature extraction condition adjustment processing unit 321 . The camera condition adjustment processing unit 311 is capable of performing camera condition adjustment processing, and adjusts the camera condition, which is a condition related to the camera 10, according to the operation of the user. The photographing target condition adjustment processing unit 312 is capable of executing photographing target condition adjustment processing, and adjusts the photographing target condition, which is a condition related to the photographing target, according to the user's operation. Furthermore, the feature extraction condition adjustment processing unit 321 is capable of executing the feature extraction condition adjustment process, and adjusts the feature extraction condition, which is a condition related to the feature point extraction process, according to the user's operation.
若根據此，則使用者可操作攝影機器條件調整處理部311、攝影對象條件調整處理部312及特徵抽出條件調整處理部321的至少1個，設定各種條件或參數，亦即可進行調諧。其結果，例如即使是實施在製造工程途中表面模樣變化的加工之類者，亦即在畫像的登錄時及識別時表面模樣變化的情況，也可彈性地對應，可高速精度佳識別。Based on this, the user can operate at least one of the imaging device condition adjustment processing unit 311 , the imaging target condition adjustment processing unit 312 , and the feature extraction condition adjustment processing unit 321 to set various conditions or parameters, that is, to perform tuning. As a result, for example, even if the surface pattern changes during the manufacturing process, that is, when the image is registered or recognized, the surface pattern can be flexibly handled, and high-speed and high-accuracy recognition can be achieved.
個體識別系統1是具備廣域特徵量字典產生處理部251。廣域特徵量字典產生處理部251是可實現廣域特徵量字典產生處理，產生使字典資訊持有從預先取得的複數的學習用畫像得到的各局部特徵量群的各個的代表值之廣域特徵量字典。而且，廣域特徵量計算處理是包含：以投票給與各局部特徵量相關性高的字典資訊而得到的柱狀圖作為廣域特徵量計算之處理。又，廣域特徵量計算處理是亦可包含：僅代表值的數量計算將局部特徵量與相關性高的字典資訊即代表值的差分向量總和後的向量，以連接該等的向量作為廣域特徵量計算之處理。The individual identification system 1 includes a wide-area feature amount dictionary generation processing unit 251 . The wide-area feature value dictionary generation processing unit 251 is a wide-area feature value dictionary generation process capable of generating a wide-area feature value dictionary in which the representative value of each local feature value group obtained from a complex number of learning images obtained in advance is included in the dictionary information. feature dictionary. Further, the wide-area feature amount calculation process includes a process of calculating a wide-area feature amount using a histogram obtained by voting for dictionary information having a high correlation with each local feature amount. In addition, the wide-area feature amount calculation process may include: calculating a vector obtained by summing the local feature amounts and the dictionary information with high correlation, that is, the difference vectors of the representative values only by the number of representative values, and connecting these vectors as a wide-area vector. Processing of feature quantity calculation.
亦即，依被照體的種類或狀態，以怎樣的地方作為特徵點抽出會改變，因此只單純地分類從畫像取得的局部特徵量來計算廣域特徵量，是有廣域特徵量的性能低，亦即識別能力低的可能性。相對於此，若根據本構成，則藉由使用具有適於被照體的字典資訊之廣域特徵量字典，即使被照體的種類或狀態改變，也可取得性能佳的廣域特徵量，其結果，可提升鎖定處理的精度。That is, depending on the type or state of the subject, where the feature points are extracted varies, so only the local feature values obtained from the portrait are simply classified to calculate the wide-area feature value, which has the performance of the wide-area feature value. Low, that is, the probability of low recognition ability. On the other hand, according to the present configuration, by using a wide-area feature quantity dictionary having dictionary information suitable for the subject, even if the type or state of the subject is changed, a wide-area feature quantity with good performance can be obtained. As a result, the precision of the locking process can be improved.
並且，在本構成中，廣域特徵量字典產生處理是包含以局部特徵量群的重心作為代表值設定於字典資訊的處理。局部特徵量群的重心是代表其局部特徵量群者。因此，藉由以局部特徵量群的重心作為代表值設定於字典資訊，可取得性能佳的廣域特徵量字典。In addition, in this configuration, the wide-area feature amount dictionary generation process includes a process of setting the center of gravity of the local feature amount group as a representative value in the dictionary information. The center of gravity of the local feature quantity group is what represents its local feature quantity group. Therefore, by setting the center of gravity of the local feature group as a representative value in the dictionary information, a wide-area feature dictionary with good performance can be obtained.
個體識別系統1是具備畫像取得處理部212及特徵點抽出處理部221，更具備特徵點顯示處理部331。特徵點顯示處理部331是可實行特徵點顯示處理，例如圖21所示般，將藉由特徵點抽出處理所抽出的特徵點重疊於畫像而顯示於顯示裝置12。The individual identification system 1 includes an image acquisition processing unit 212 and a feature point extraction processing unit 221 , and further includes a feature point display processing unit 331 . The feature point display processing unit 331 is capable of executing feature point display processing, for example, as shown in FIG. 21 , and displays the feature points extracted by the feature point extraction processing on the image to be displayed on the display device 12 .
個體識別系統1是具備攝影機器條件調整處理部311、攝影對象條件調整處理部312及特徵抽出條件調整處理部321之中至少任一個。攝影機器條件調整處理部311是可實行攝影機器條件調整處理，根據使用者的操作調整有關攝影機器10的條件即攝影機器條件。攝影對象條件調整處理部312是可實行攝影對象條件調整處理，根據使用者的操作調整有關攝影對象的條件即攝影對象條件。而且，特徵抽出條件調整處理部321是可實行特徵抽出條件調整處理，根據使用者的操作調整有關特徵點抽出處理的條件即特徵抽出條件。The individual identification system 1 includes at least any one of a photographing device condition adjustment processing unit 311 , a photographing object condition adjustment processing unit 312 , and a feature extraction condition adjustment processing unit 321 . The camera condition adjustment processing unit 311 is capable of performing camera condition adjustment processing, and adjusts the camera condition, which is a condition related to the camera 10, according to the operation of the user. The photographing target condition adjustment processing unit 312 is capable of executing photographing target condition adjustment processing, and adjusts the photographing target condition, which is a condition related to the photographing target, according to the user's operation. Furthermore, the feature extraction condition adjustment processing unit 321 is capable of executing the feature extraction condition adjustment process, and adjusts the feature extraction condition, which is a condition related to the feature point extraction process, according to the user's operation.
例如在生產現場使用個體識別系統1時，成為識別的對象的零件是有製造工程中的歷時變化或熱加工等而其表面模樣的色調等變化者。如此的實施在製造工程途中表面模樣的色調等變化的加工的零件中，其加工的前後，亦即在畫像的登錄時及識別時被抽出的特徵點的位置或數量及局部特徵量有大幅度變化的可能性，其結果，識別精度會降低。For example, when the individual identification system 1 is used at a production site, the parts to be identified are those that have changes in the color tone of their surface patterns due to changes over time in the manufacturing process, thermal processing, and the like. In the parts subjected to processing such that the color tone of the surface pattern changes in the middle of the manufacturing process, the position or number of feature points and local feature quantities extracted before and after processing, that is, at the time of image registration and recognition, are greatly increased. The possibility of change, as a result, the recognition accuracy will decrease.
對於此，若根據本構成，則由於在特徵點抽出處理抽出的特徵點會被重疊於畫像而顯示於顯示裝置12，因此使用者可目視確認畫像之中的何處作為特徵點被抽出。然後，使用者可操作特徵抽出條件調整處理部321、攝影對象條件調整處理部312及特徵抽出條件調整處理部321的至少1個，使可針對表面模樣即使經過製造工程也不易變化之處重點地抽出特徵點，邊確認被抽出的特徵點的位置，邊設定各種條件或參數，亦即進行調諧。其結果，即使是實施在製造工程途中表面模樣變化的加工之類者，亦即在畫像的登錄時及識別時表面模樣變化的情況，也可取得能精度佳識別的良好效果。In contrast, according to this configuration, the feature points extracted by the feature point extraction process are displayed on the display device 12 by being superimposed on the image, so the user can visually confirm where in the image the feature points are extracted. Then, the user can operate at least one of the feature extraction condition adjustment processing unit 321, the imaging target condition adjustment processing unit 312, and the feature extraction condition adjustment processing unit 321, so that the surface pattern is not easily changed even after the manufacturing process. A feature point is extracted, and various conditions or parameters are set while confirming the position of the extracted feature point, that is, tuning is performed. As a result, even in cases where the surface pattern changes during the manufacturing process, that is, the surface pattern changes during registration and recognition of an image, it is possible to obtain an excellent effect of being able to recognize with high accuracy.
個體識別系統1是具備局部特徵量計算處理部222及局部特徵量群分類處理部223。而且，特徵點顯示處理是更包含：如圖21(A)所示般，以按局部特徵量所屬的每個局部特徵量群而異的形態，例如使顏色或大小、形狀不同的形態來將特徵點顯示於顯示裝置12之處理。The individual identification system 1 includes a local feature value calculation processing unit 222 and a local feature value group classification processing unit 223 . Furthermore, the feature point display processing further includes, as shown in FIG. 21(A) , in a form that differs for each local feature quantity group to which the local feature quantity belongs, for example, a form that differs in color, size, and shape. The feature points are displayed in the processing of the display device 12 .
若根據此，則使用者可一眼就能掌握什麼樣的特徵點存在於畫像中的何處。藉此，邊確認被抽出的特徵點的位置，邊設定各種條件或參數的作業更容易。其結果，可更精度佳識別有關實施在製造工程途中表面模樣變化的加工之類者。According to this, the user can grasp what kind of feature points exist and where in the image at a glance. This makes it easier to set various conditions and parameters while confirming the positions of the extracted feature points. As a result, it is possible to more accurately recognize the processing or the like that is performed on the surface pattern change in the middle of the manufacturing process.
個體識別系統1是具備廣域特徵量計算處理部224，更具備廣域特徵量計算統計資訊顯示處理部332。廣域特徵量計算統計資訊顯示處理部332是可實行廣域特徵量計算統計資訊顯示處理，如圖21(B)所示般，在藉由廣域特徵量字典產生處理的實行之廣域特徵量的計算過程中，將在局部特徵量群分類處理的過程取得的廣域特徵量計算統計資訊G顯示於顯示裝置12。The individual identification system 1 includes a wide-area feature amount calculation processing unit 224 , and further includes a wide-area feature amount calculation statistical information display processing unit 332 . The wide-area feature amount calculation statistical information display processing unit 332 is capable of executing the wide-area feature amount calculation statistical information display process, as shown in FIG. 21(B), in the implementation of the wide-area feature amount dictionary generation process by the wide-area feature amount In the calculation process of the quantity, the wide-area feature quantity calculation statistical information G obtained in the process of the local feature quantity group classification process is displayed on the display device 12 .
若根據此，則使用者容易掌握取得了怎樣的廣域特徵量。亦即，藉由顯示有多少投票給域特徵量字典的哪個字典資訊之資訊，使用者可掌握該域特徵量字典所具有的字典資訊的選擇者的偏向。藉此，邊看被顯示的廣域特徵量計算統計資訊G，邊設定有關廣域特徵量的各種條件或參數之作業更容易。其結果，可更精度佳識別有關實施在製造工程途中表面模樣變化的加工之類者。According to this, the user can easily grasp what kind of wide-area feature amount has been acquired. That is, by displaying the information of how many votes for which dictionary information of the domain feature quantity dictionary, the user can grasp the preference of the selector of the dictionary information possessed by the domain feature quantity dictionary. This makes it easier to set various conditions and parameters related to the wide-area feature amount while viewing the displayed wide-area feature amount calculation statistic information G. As a result, it is possible to more accurately recognize the processing or the like that is performed on the surface pattern change in the middle of the manufacturing process.
個體識別系統1是更具備廣域特徵量字典產生處理部251。廣域特徵量字典產生處理部251是可實行廣域特徵量字典產生處理，產生使字典資訊持有從預先取得的複數的學習用畫像得到的各局部特徵量群的各個的代表值之廣域特徵量字典。而且，廣域特徵量計算統計資訊顯示處理是包含：如圖13所示般，以在廣域特徵量的計算過程中投票給與各局部特徵量相關性高的字典資訊d而得到的柱狀圖或相關值的統計量的得分作為廣域特徵量計算統計資訊G，如圖21般顯示於顯示裝置12之處理。The individual identification system 1 further includes a wide-area feature amount dictionary generation processing unit 251 . The wide-area feature amount dictionary generation processing unit 251 is capable of executing a wide-area feature amount dictionary generation process, and generates a wide area in which the dictionary information contains the representative value of each local feature amount group obtained from the plural learning images obtained in advance. feature dictionary. Further, the wide-area feature amount calculation statistical information display process includes, as shown in FIG. 13 , a columnar shape obtained by voting for dictionary information d having a high correlation with each local feature amount in the calculation process of the wide-area feature amount The score of the statistical quantity of the graph or the correlation value is calculated as the wide-area feature quantity, and the statistical information G is displayed on the display device 12 as shown in FIG. 21 .
若根據此，則使用者容易視覺地且直覺地掌握廣域特徵量。藉此，邊看被顯示的廣域特徵量計算統計資訊G，邊設定有關廣域特徵量的各種條件或參數之作業更容易。其結果，可更精度佳識別有關實施在製造工程途中表面模樣變化的加工之類者。According to this, the user can easily grasp the wide-area feature quantity visually and intuitively. This makes it easier to set various conditions and parameters related to the wide-area feature amount while viewing the displayed wide-area feature amount calculation statistic information G. As a result, it is possible to more accurately recognize the processing or the like that is performed on the surface pattern change in the middle of the manufacturing process.
個體識別系統1是更具備廣域特徵量字典設定處理部322。廣域特徵量字典設定處理部322是可實行廣域特徵量字典設定處理，按每個廣域特徵量計算統計資訊來計算得分，自動設定其得分高者作為用在局部特徵量群分類處理的廣域特徵量字典。The individual identification system 1 further includes a wide-area feature amount dictionary setting processing unit 322 . The wide-area feature value dictionary setting processing unit 322 is capable of executing a wide-area feature value dictionary setting process, calculating statistical information for each wide-area feature value to calculate a score, and automatically setting the one with the highest score as the one used in the local feature value group classification process. A dictionary of wide-area feature quantities.
若根據此，則由於作為適當的廣域特徵量字典是自動設定，因此使用者不須逐一手動調整條件或參數，其結果，使用者方便性會提升。According to this, since the appropriate wide-area feature amount dictionary is automatically set, the user does not need to manually adjust the conditions or parameters one by one, and as a result, the user convenience is improved.
例如在工廠等使用本構成的個體識別系統1時，假想登錄畫像的攝影及識別畫像的攝影是在不同的場所進行。此情況，在工廠內產生的粉塵或工廠內的照明、外光等的自然光、周圍的雜訊源等會成為干擾的要因，登錄畫像的攝影與識別畫像的攝影的攝影環境有大幅度改變的可能性。於是，被抽出的特徵點或局部特徵量等會改變，恐有識別性能降低之虞。For example, when the individual identification system 1 of the present configuration is used in a factory or the like, the imaging of the virtual registration portrait and the imaging of the identification portrait are performed in different places. In this case, the dust generated in the factory, the lighting in the factory, natural light such as outside light, and surrounding noise sources will be the cause of interference, and the shooting environment of the registration image and the identification image will change greatly. possibility. As a result, the extracted feature points, local feature quantities, etc. may change, and there is a possibility that the recognition performance may be lowered.
於是，本構成的個體識別系統1是具備畫像取得處理部212、特徵點抽出處理部221、局部特徵量計算處理部222、攝影對象條件調整處理部312及攝影機器條件調整處理部311。藉此，使用者是可操作攝影對象條件調整處理部312及攝影機器條件調整處理部311，來調整適於特徵點的抽出及局部特徵量的計算之攝影對象條件及攝影機器條件。因此，藉由按照環境的變化來設定適當的條件，可使干擾的影響減低，可使所欲作為特徵的模樣適當地浮起。其結果，可使識別精度提升。Therefore, the individual recognition system 1 of the present configuration includes an image acquisition processing unit 212 , a feature point extraction processing unit 221 , a local feature amount calculation processing unit 222 , a photographing target condition adjustment processing unit 312 , and a photographing equipment condition adjustment processing unit 311 . Thereby, the user can operate the imaging object condition adjustment processing unit 312 and the imaging device condition adjustment processing unit 311 to adjust the imaging object conditions and imaging device conditions suitable for extraction of feature points and calculation of local feature amounts. Therefore, by setting appropriate conditions according to changes in the environment, the influence of disturbance can be reduced, and the desired feature pattern can be appropriately raised. As a result, the recognition accuracy can be improved.
個體識別裝置20是更具備總括條件產生處理部313。總括條件產生處理部313是可實行總括條件產生處理，總括產生預先被設定的範圍內的攝影對象條件與預先被設定的範圍內的攝影機器條件的組合。而且，畫像取得處理是包含：自動取得以在總括條件產生處理產生的全部的條件攝影後的畫像之處理。The individual identification device 20 further includes a collective condition generation processing unit 313 . The collective condition generation processing unit 313 is capable of executing collective condition generation processing, and collectively generates a combination of photographing target conditions within a preset range and photographing device conditions within a preset range. Further, the image acquisition process includes a process of automatically acquiring an image captured under all the conditions generated in the collective condition generation process.
若根據此，則可自動攝影取得改變了攝影對象條件及攝影機器條件的各者之多數的畫像。因此，在改變攝影對象條件及攝影機器條件的各者來尋找最適的條件的組合時，不需要以使用者的手工作業來產生條件的組合、攝影。其結果，可減低調整攝影對象條件及攝影機器條件的條件之工夫。而且，可簡單地設定性能佳的條件，因此識別性能的提升也可謀求。According to this, it is possible to automatically capture and acquire images in which most of the conditions of the subject to be photographed and the conditions of the photographing device have been changed. Therefore, when searching for an optimum combination of conditions by changing each of the conditions of the photographic object and the conditions of the photographing machine, it is not necessary to generate the combination of conditions and photograph by the user's manual work. As a result, it is possible to reduce the time and effort required to adjust the conditions of the photographic object and the conditions of the photographing equipment. In addition, the conditions for good performance can be easily set, so that the recognition performance can also be improved.
個體識別裝置20是更具備最適條件提示處理部314。最適條件提示處理部314是可實行最適條件提示處理，以在總括條件產生處理產生的全部的條件來攝影後的登錄畫像及識別畫像之中，提示攝取了相同的被照體的前述登錄畫像與前述識別畫像的對應點數多的攝影對象條件和攝影機器條件的組合作為最適條件。若根據此，則使用者可簡單地取得特徵點的數量多的最適的條件設定，因此可更減低調整攝影對象條件及攝影機器條件的條件之工夫。而且，可簡單地設定攝影對象條件及攝影機器條件的最適的組合，因此識別性能的提升也可謀求。The individual identification device 20 further includes an optimum condition presentation processing unit 314 . The optimum condition presentation processing unit 314 is capable of performing optimum condition presentation processing to present the registration image and the identification image that have captured the same subject among the registration images and recognition images captured under all the conditions generated by the collective condition generation processing. The combination of the conditions of the photographing object and the conditions of the photographing equipment in which the number of corresponding points of the recognition image is large is regarded as the optimum condition. According to this, the user can easily obtain the optimum condition setting with a large number of feature points, and therefore, the time and effort to adjust the conditions of the photographic object conditions and the conditions of the photographing device can be further reduced. Furthermore, since it is possible to easily set the optimum combination of the conditions of the photographic object and the conditions of the photographing device, it is possible to improve the recognition performance.
最適條件提示處理是包含：預先複數準備有關特徵點抽出處理的條件即特徵抽出條件，針對以在總括條件產生處理產生的全部的條件來攝取了相同的被照體之登錄畫像及識別畫像，利用以預先準備的複數的特徵抽出條件來實行特徵點抽出處理而取得的特徵點來對照時的對應點數成為最多的條件的組合提示之處理。
若根據此，則使用者是除攝影對象條件及攝影機器條件的條件之外，還可簡單地取得與有關特徵點抽出處理的特徵抽出條件的組合之中最適的條件。藉此，可更減低調整攝影對象條件、攝影機器條件及特徵抽出條件的工夫。而且，可簡單地設定攝影對象條件、攝影機器條件及特徵抽出條件的最適的組合，因此可謀求識別性能的進一步的提升。The optimum condition presentation process includes preparing a plurality of conditions for the feature point extraction process, that is, feature extraction conditions, and using the registration images and recognition images of the same subject captured under all the conditions generated in the collective condition generation process. A process of presenting a combination of conditions that correspond to the largest number of corresponding points when comparing the feature points obtained by executing the feature point extraction process with the plurality of feature extraction conditions prepared in advance.
According to this, the user can easily obtain the optimum condition in combination with the feature extraction condition related to the feature point extraction process, in addition to the conditions of the photographing object and the photographing device. Thereby, the labor for adjusting the conditions of the photographic object, the conditions of the photographing equipment, and the conditions of feature extraction can be further reduced. Furthermore, since the optimum combination of the photographing target conditions, the photographing device conditions, and the feature extraction conditions can be easily set, it is possible to further improve the recognition performance.
例如在工業製品的零件中，與人的臉或指紋不同，依照所欲對照識別的對象的種類或材質、加工方法，被顯現於其表面的模樣的特徵也各式各樣。因此，需要依照所欲識別的對象的種類來將用在對照識別的算法的參數調整成適當者。但，參數的項目有多數，使用者以手動來從其全部的組合之中尋找最適的組合是非常費工夫及時間。For example, in parts of industrial products, unlike a human face or a fingerprint, the features of the pattern appearing on the surface vary depending on the type, material, and processing method of the object to be compared and identified. Therefore, it is necessary to adjust the parameters of the algorithm used for the comparison recognition to be appropriate according to the type of the object to be recognized. However, there are many parameter items, and it takes a lot of time and effort for the user to manually search for the optimum combination from all the combinations.
於是，個體識別系統1是具備特徵點抽出處理部221及局部特徵量計算處理部222。又，個體識別系統1是更具備對照處理部262、參數組產生處理部324及參數組設定處理部325。對照處理部262是可實行對照處理，比較登錄畫像與識別畫像的局部特徵量，取得特徵點的對應點數而對照登錄畫像與識別畫像。參數組產生處理部324是可實行參數組產生處理，針對用以設定對照處理的條件之參數，組合被設定成各參數所定的設定範圍內的設定值之各參數來產生複數的參數組。參數組設定處理部325是可實行根據使用者的操作來設定參數組的參數組設定處理。Therefore, the individual identification system 1 includes a feature point extraction processing unit 221 and a local feature amount calculation processing unit 222 . In addition, the individual identification system 1 further includes a comparison processing unit 262 , a parameter group generation processing unit 324 , and a parameter group setting processing unit 325 . The collation processing unit 262 can perform collation processing, compare the local feature quantities of the registered image and the recognized image, obtain the number of corresponding points of the feature points, and compare the registered image and the recognized image. The parameter group generation processing unit 324 is capable of executing a parameter group generation process, and generates a plurality of parameter groups by combining the parameters set to the setting values within the setting range specified by each parameter with respect to the parameters for setting the conditions of the comparison processing. The parameter group setting processing unit 325 is capable of executing a parameter group setting process for setting a parameter group according to a user's operation.
對照處理是包含：利用在參數組產生處理產生的各參數組來實行對照處理之處理。而且，個體識別系統1是更具備對照結果顯示處理部333。對照結果顯示處理部333是可實行對照結果顯示處理，將利用各參數組來實行對照處理時的對照結果匯集而顯示於顯示裝置12。The collation process is a process including executing the collation process using each parameter set generated in the parameter set generating process. Furthermore, the individual identification system 1 further includes a comparison result display processing unit 333 . The collation result display processing unit 333 is capable of executing the collation result display process, and displays the collation results obtained when the collation process is executed using each parameter group on the display device 12 in aggregate.
若根據此，則參數的組合即參數組會自動產生，利用其各參數組來進行對照處理。因此，使用者不須逐一手工作業作成參數的組合進行對照。而且，其對照結果是被彙整顯示於顯示裝置12。因此，使用者可藉由看被顯示於顯示裝置12的結果來確認各參數組的性能。然後，使用者是只要看被顯示於顯示裝置12的結果來選擇適當的參數組即可。藉此，不論所欲對照識別的對象的種類，可簡單地不費工夫精度佳調整用在對照識別的算法的參數。According to this, a combination of parameters, that is, a parameter group will be automatically generated, and each parameter group will be used for comparison processing. Therefore, the user does not need to manually create parameter combinations one by one for comparison. Furthermore, the comparison result is displayed on the display device 12 in a consolidated manner. Therefore, the user can confirm the performance of each parameter set by looking at the results displayed on the display device 12 . Then, the user only needs to select an appropriate parameter group by looking at the result displayed on the display device 12 . In this way, regardless of the type of the object to be compared and recognized, the parameters of the algorithm used for the comparison and recognition can be easily and accurately adjusted without much effort.
對照結果顯示處理是包含：如圖22所示般，將利用各參數組來實行對照處理的對照結果圖表化而顯示的處理。容易視覺地且直覺地掌握複數的參數組之中性能佳者。藉此，可簡單地進行精度佳的參數的調整作業。The comparison result display process includes, as shown in FIG. 22 , a process of graphing and displaying the comparison result obtained by performing the comparison process using each parameter group. It is easy to visually and intuitively grasp the best performers among complex parameter groups. Thereby, the adjustment operation of the parameter with high precision can be performed easily.
個體識別系統1是更具備評價用畫像登錄處理部232。評價用畫像登錄處理部232是可實行評價用畫像登錄處理，該評價用畫像登錄處理是以登錄畫像群及識別畫像作為對照處理的評價用的畫像，在互相建立關聯的狀態下登錄於評價用畫像登錄部242，該登錄畫像群是包含攝取了各個不同的被照體的複數的登錄畫像，該識別畫像是以和登錄畫像群之中的1個相同的被照體且不同的條件來攝影的識別畫像。The individual identification system 1 further includes an evaluation image registration processing unit 232 . The evaluation portrait registration processing unit 232 is capable of executing evaluation portrait registration processing. The evaluation portrait registration process is to register the portrait group and the identification portrait as an evaluation portrait in a state of being associated with each other and register them for evaluation. The image registration unit 242, the registered image group includes a plurality of registered images in which different subjects are captured, and the identification image is photographed under different conditions from the same subject as one of the registered image groups identification image.
若根據此，則使用者在進行參數組的評價時，可不須逐一攝取畫像。其結果，可簡單地進行參數組的評價。又，若根據此，則攝取了相同的被照體的登錄畫像及識別畫像是在互相建立關聯的狀態下被登錄。因此，在利用登錄畫像及識別畫像來實行對照處理時，可事前分明登錄畫像與識別畫像的對應關係，亦即攝取了相同的被照體者或攝取了不同的被照體者。因此，藉由利用對應關係分明的畫像，可明確在其對照處理中取得的結果，使用者可容易評價用在對照處理的參數組。According to this, the user does not need to capture the images one by one when evaluating the parameter set. As a result, the parameter group evaluation can be performed easily. In addition, according to this, the registered portrait and the identification portrait that have captured the same subject are registered in a state in which they are associated with each other. Therefore, when the comparison processing is performed using the registered portrait and the identification portrait, the correspondence between the registered portrait and the identification portrait, that is, the same subject or a different subject can be identified in advance. Therefore, by using the images with clear correspondences, the results obtained in the comparison processing can be made clear, and the user can easily evaluate the parameter set used in the comparison processing.
亦即，若利用攝取了相同的被照體的登錄畫像及識別畫像來實行對照處理時對應點數多，且利用攝取了不同的被照體的登錄畫像及識別畫像來實行對照處理時對應點數少，則使用者可評價成用在該對照處理的參數組的性能高。相反的，若利用攝取了相同的被照體的登錄畫像及識別畫像來實行對照處理時對應點數少，或利用攝取了不同的被照體的登錄畫像及識別畫像來實行對照處理時對應點數少，則使用者可評價成用在該對照處理的參數組的性能低。That is, if the comparison processing is performed using the registration image and the recognition image of the same subject, there are many corresponding points, and the corresponding points are performed using the registration image and the recognition image of different subjects. If the number is small, the user can evaluate the performance of the parameter set used in the control treatment as high. Conversely, if the comparison processing is performed using the registration image and the recognition image of the same subject, the number of corresponding points is small, or if the registration image and the recognition image of different subjects are used for the comparison processing. If the number is small, the user may assess that the performance of the parameter set used in the control treatment is low.
對照處理是包含：對照在登錄畫像群中所含的1個以上的登錄畫像與識別畫像之處理。而且，對照結果顯示處理是包含：如圖23所示般，將根據在對照處理計算的對應點數之對應點數評價資訊顯示於顯示裝置12的處理。若根據此，則使用者藉由看被顯示於顯示裝置12的對應點數評價資訊，可簡單地評價該參數組的性能，藉此可按照所欲對照識別的對象的種類來選擇適當的參數組。其結果，不管是怎樣的種類的對象也可發揮高的對照識別性能。The collation process is a process including collation of one or more registered portraits included in the registered portrait group with the identification portrait. Furthermore, the collation result display process includes a process of displaying on the display device 12 the corresponding point evaluation information based on the corresponding points calculated in the collation process, as shown in FIG. 23 . According to this, the user can easily evaluate the performance of the parameter set by looking at the corresponding point evaluation information displayed on the display device 12, and can select appropriate parameters according to the type of objects to be compared and recognized. Group. As a result, regardless of the type of object, high collation recognition performance can be exhibited.
個體識別裝置20是更具備特徵抽出條件調整處理部321。特徵抽出條件調整處理部321是可實行特徵抽出條件調整處理。特徵抽出條件調整處理是包含計算正對應點數及負對應點數之處理，該正對應點數是登錄畫像群之中攝取了和識別畫像的被照體相同的被照體的登錄畫像與識別畫像的對應點數，負對應點數是登錄畫像群之中攝取了和識別畫像的被照體不同的被照體的登錄畫像與識別畫像的對應點數。而且，特徵抽出條件調整處理是包含：針對局部特徵量計算處理所致的特徵抽出條件，調整正對應點數與負對應點數的比能成為最大的處理。The individual identification device 20 further includes a feature extraction condition adjustment processing unit 321 . The feature extraction condition adjustment processing unit 321 is capable of executing feature extraction condition adjustment processing. The feature extraction condition adjustment processing includes the processing of calculating the number of positive correspondence points and the number of negative correspondence points, the positive correspondence points being the registered portraits and identifications of the subjects in the registered portrait group that have captured the same subjects as those of the identification portraits. The corresponding points of the portraits, and the negative corresponding points are the corresponding points of the registered portraits and the identification portraits in the registered portrait group that have captured subjects different from those of the identification portraits. Furthermore, the feature extraction condition adjustment process includes a process of adjusting the ratio of the number of positive correspondence points to the number of negative correspondence points to maximize the feature extraction condition by the local feature amount calculation process.
此情況，特徵抽出條件調整處理部321是以正對應點數與負對應點數的比成為最大的方式，尋找正對應點數多且負對應點數少的特徵抽出條件。所謂正對應點數多，是意思攝取了相同的被照體的畫像間的對應點數多。這是意思在拍攝了相同的被照體的畫像間的對照處理中，識別成畫像中所含的被照體為相同者的可能性高，亦即拍攝了相同的被照體的畫像間的識別性能高。又，所謂負對應點數少，是意思拍攝了不同的被照體的畫像間的對應點數少。這是意思在拍攝了不同的被照體的畫像間的對照處理中，識別成畫像中所含的被照體為不同者的可能性高，亦即拍攝了不同的被照體的畫像間的識別性能高。In this case, the feature extraction condition adjustment processing unit 321 searches for a feature extraction condition with a large number of positive correspondence points and a small number of negative correspondence points so that the ratio of the number of positive correspondence points to the number of negative correspondence points becomes the largest. That the number of positive correspondence points is large means that the number of correspondence points between images that have captured the same subject is large. This means that, in the comparison processing between portraits in which the same subject was photographed, there is a high probability that the subjects included in the portraits are the same, that is, the comparison between portraits in which the same subject was photographed High recognition performance. In addition, that the number of negative correspondence points is small means that the number of correspondence points between images of different subjects is small. This means that in the comparison processing between portraits in which different subjects are photographed, it is highly likely that the subjects included in the portraits are different people, that is, the difference between portraits in which different subjects are photographed is recognized. High recognition performance.
如此，特徵抽出條件調整處理部321是自動調整特徵抽出條件，使拍攝了相同的被照體的畫像間及拍攝了不同的被照體的畫像間的對照識別性能變高，藉此可邊取得高的對照識別性能，邊省去花在調整的使用者的工夫。In this way, the feature extraction condition adjustment processing unit 321 automatically adjusts the feature extraction conditions so as to improve the comparison and recognition performance between images of the same subject and images of different subjects. The high contrast recognition performance saves the user's effort in adjustment.
個體識別系統1是具備局部特徵量群分類處理部223及廣域特徵量計算處理部224。而且，個體識別系統1是更具備廣域特徵量相關係數計算處理部323。廣域特徵量相關係數計算處理部323可實行廣域特徵量相關係數計算處理，比較各登錄畫像的廣域特徵量與識別畫像的廣域特徵量，計算表示其相關關係的廣域特徵量相關係數。而且，對照結果顯示處理是包含：將根據廣域特徵量相關係數的資訊顯示於顯示裝置12之處理。The individual identification system 1 includes a local feature value group classification processing unit 223 and a wide area feature value calculation processing unit 224 . Furthermore, the individual identification system 1 further includes a wide-area feature quantity correlation coefficient calculation processing unit 323 . The wide-area feature amount correlation coefficient calculation processing unit 323 can perform a wide-area feature amount correlation coefficient calculation process, compare the wide-area feature amount of each registered profile with the wide-area feature amount of the recognition profile, and calculate the wide-area feature amount correlation indicating the correlation. coefficient. Further, the comparison result display process includes a process of displaying the information based on the correlation coefficient of the wide-area feature amount on the display device 12 .
廣域特徵量相關係數是成為表示登錄畫像與識別畫像的廣域特徵量的一致度亦即性能的1個的指標。藉由在顯示裝置12顯示根據廣域特徵量相關係數的資訊，使用者可容易掌握廣域特徵量的性能。因此，使用者可在顯示裝置12以根據廣域特徵量相關係數的資訊作為1個目標進行各種條件或參數的調整作業，所以作業變更容易。The correlation coefficient of the wide-area feature quantity is an index indicating the degree of agreement between the registration profile and the wide-area feature quantity of the recognition profile, that is, performance. By displaying the information according to the correlation coefficient of the wide-area feature amount on the display device 12, the user can easily grasp the performance of the wide-area feature amount. Therefore, the user can perform the adjustment operation of various conditions or parameters on the display device 12 with the information based on the correlation coefficient of the wide-area feature amount as a target, so that the operation can be easily changed.
個體識別系統1是更具備廣域特徵量字典登錄部245及廣域特徵量字典設定處理部322。廣域特徵量字典設定處理部322是可實行廣域特徵量字典設定處理。廣域特徵量字典設定處理是包含：選擇被登錄於廣域特徵量字典登錄部245的複數的廣域特徵量字典之中，可根據廣域特徵量相關係數，將攝取了相同的被照體的登錄畫像與識別畫像判斷成相同，將攝取了不同的被照體的登錄畫像與識別畫像判斷成不相同的廣域特徵量字典，設定成用在局部特徵量群分類處理的廣域特徵量字典之處理。The individual identification system 1 further includes a wide area feature amount dictionary registration unit 245 and a wide area feature amount dictionary setting processing unit 322 . The wide-area feature amount dictionary setting processing unit 322 is capable of executing the wide-area feature amount dictionary setting process. The wide-area feature value dictionary setting process includes selecting a plurality of wide-area feature value dictionaries registered in the wide-area feature value dictionary registration unit 245, and then, based on the wide-area feature value correlation coefficient, the same subject can be captured. The registration image and the recognition image are determined to be the same, and the wide-area feature quantity dictionary in which the registration image and the recognition image of different subjects are judged to be different is set as the wide-area feature quantity used in the classification processing of the local feature quantity group. Handling of dictionaries.
廣域特徵量相關係數是如上述般，成為表示登錄畫像與識別畫像的廣域特徵量的一致度的1個的指標。意思2個的畫像間的廣域特徵量的相關性越大，在兩者的畫像含有相同的被照體的可能性越高。因此，攝取了相同的被照體的畫像間的廣域特徵量的相關性大，是意思針對攝取了該相同的被照體的2個的畫像計算的廣域特徵量的一致度高。亦即，這是意思在攝取了相同的被照體的畫像間其廣域特徵量字典的性能高。The wide-area feature value correlation coefficient is, as described above, an index that represents the degree of agreement between the registered image and the wide-area feature value of the recognized image. This means that the greater the correlation between the wide-area feature quantities between the two images, the higher the possibility that the two images contain the same subject. Therefore, the high correlation of the wide-area feature amounts between the images that capture the same subject means that the degree of agreement of the wide-area feature amounts calculated for the two images that capture the same subject is high. That is, this means that the performance of the wide-area feature-quantity dictionary is high among the images that have captured the same subject.
另一方面，2個的畫像間的廣域特徵量的相關性越小，在兩者的畫像中所含的被照體不同的可能性越高。因此，攝取了不同的被照體的畫像間的廣域特徵量的相關性小，是意思針對攝取了該不同的被照體的畫像，使用其廣域特徵量字典來計算的廣域特徵量的一致度低。亦即，這是意思在攝取了不同的被照體的畫像間其廣域特徵量字典的性能高。On the other hand, the smaller the correlation between the wide-area feature amounts between the two portraits, the higher the possibility that the subjects included in the two portraits are different. Therefore, the correlation between the wide-area feature quantities between the images that have captured different subjects is small, which means that the wide-area feature quantities are calculated using the wide-area feature quantity dictionary for the images that have captured the different subjects. The consistency is low. That is, this means that the performance of the wide-area feature-quantity dictionary is high between images of different subjects.
若根據本構成，則可藉由廣域特徵量字典設定處理來自動設定在拍攝了相同的被照體的畫像間的性能高，且在拍攝了不同的被照體的畫像間的性能也高的廣域特徵量字典。因此，省去使用者的調整的工夫，且可取得高的對照識別性能。According to this configuration, the wide-area feature value dictionary setting process can automatically set high performance between images of the same subject and high performance between images of different subjects. The wide-area feature dictionary of . Therefore, the user's effort for adjustment is omitted, and high contrast recognition performance can be obtained.
廣域特徵量字典設定處理是亦可包括：當2個的廣域特徵量為向量時，分別將絶對值正規化成1，取內積的值，針對該內積的值接近1者判斷成同一性高，針對該內積的值接近-1者判斷成同一性低的處理。又，廣域特徵量相關係數是可設為2個的廣域特徵量的向量間的距離的值。此情況，廣域特徵量字典設定處理是包括：若2個的廣域特徵量的向量間的距離的值接近0，則判斷成同一性高，值大時則判斷成同一性低的處理。若根據該等的構成，則可定量地評價廣域特徵量字典的性能。The wide-area feature amount dictionary setting process may also include: when the two wide-area feature amounts are vectors, normalize the absolute value to 1 respectively, take the value of the inner product, and judge that the value of the inner product is close to 1 as the same If the value of the inner product is close to -1, it is judged that the identity is low. In addition, the wide-area feature quantity correlation coefficient is a value of the distance between vectors of two broad-area feature quantity vectors. In this case, the wide-area feature value dictionary setting process includes a process of determining that the identity is high when the value of the distance between two vectors of wide-area feature values is close to 0, and determining that the identity is low when the value is large. According to these configurations, the performance of the wide-area feature amount dictionary can be quantitatively evaluated.
個體識別系統1是具備：具有畫像取得處理部212、特徵點抽出處理部221及特定處理部2612的個體識別裝置20。而且，個體識別系統1是除了個體識別裝置20以外，還具備可視化裝置33、攝影條件調整裝置31及對照識別條件調整裝置32之中至少2個的裝置。The individual identification system 1 includes an individual identification device 20 including an image acquisition processing unit 212 , a feature point extraction processing unit 221 , and a specific processing unit 2612 . Furthermore, the individual identification system 1 is a device including at least two of the visualization device 33 , the imaging condition adjustment device 31 , and the comparison identification condition adjustment device 32 in addition to the individual identification device 20 .
可視化裝置33是具有特徵點顯示處理部331。特徵點顯示處理部331是可實行特徵點顯示處理，將藉由特徵點抽出處理所抽出的特徵點重疊於前述畫像而顯示於顯示裝置12。The visualization device 33 includes a feature point display processing unit 331 . The feature point display processing unit 331 is capable of executing feature point display processing, and displays on the display device 12 the feature points extracted by the feature point extraction process, superimposed on the aforementioned image.
攝影條件調整裝置31是比較預先被登錄的複數的登錄畫像之中攝取了相同的被照體的登錄畫像與識別畫像的局部特徵量，取得特徵點的對應點數，使對照登錄畫像與識別畫像的對照處理實行於個體識別裝置20，且具有攝影對象條件調整處理部312及攝影機器條件調整處理部311。攝影對象條件調整處理部312是可實行調整關於攝影對象的條件即攝影對象條件的攝影對象條件調整處理。攝影機器條件調整處理部311是可實行調整關於攝影機器10的條件即攝影機器條件的攝影機器條件調整處理。The photographing condition adjusting device 31 compares the local feature values of the registered image and the recognition image in which the same subject is captured among the registered images registered in advance, obtains the corresponding points of the feature points, and compares the registered image with the recognition image. The matching processing of the image is performed in the individual identification device 20 , and includes a photographing object condition adjustment processing unit 312 and a photographing apparatus condition adjustment processing unit 311 . The photographing target condition adjustment processing unit 312 is capable of executing photographing target condition adjustment processing that adjusts the photographing target condition, that is, the photographing target condition. The camera condition adjustment processing unit 311 is a camera condition adjustment process capable of adjusting the camera condition, which is a condition related to the camera 10 .
對照識別條件調整裝置32是具有參數組產生處理部324及參數組設定處理部325。參數組產生處理部324是可實行參數組產生處理，針對用以設定對照處理的條件的參數，組合被設定成各參數所定的設定範圍內的設定值之各參數來產生複數的參數組。參數組設定處理部325是可實行參數組設定處理，根據使用者的操作來設定前述參數組。而且，對照識別條件調整裝置32是使利用在參數組產生處理產生的各參數組來使對照處理的處理實行於對照處理部262。而且，對照識別條件調整裝置32是更具有可實行對照結果顯示處理的對照結果顯示處理部333，利用參數組來彙整對照處理的對照結果而顯示於顯示裝置12。The collation identification condition adjustment device 32 includes a parameter group generation processing unit 324 and a parameter group setting processing unit 325 . The parameter group generation processing unit 324 is capable of executing a parameter group generation process, and generates a plurality of parameter groups by combining the parameters set to the setting values within the setting range specified by the parameters for the parameters for setting the conditions of the comparison processing. The parameter group setting processing unit 325 is capable of executing a parameter group setting process, and sets the aforementioned parameter group according to the operation of the user. Furthermore, the collation identification condition adjustment device 32 executes the collation processing unit 262 in the collation process using each parameter set generated in the parameter set generation process. Further, the matching identification condition adjustment device 32 further includes a matching result display processing unit 333 capable of executing the matching result display processing, and displays the matching results of the matching processing on the display device 12 by assembling the matching results using the parameter group.
藉由個體識別系統1具備可視化裝置33，在特徵點抽出處理抽出的特徵點會被重疊於畫像而顯示於顯示裝置12，因此使用者可目視確認畫像之中的何處作為特徵點被抽出。然後，使用者可邊確認被抽出的特徵點的位置，邊設定各種條件或參數，亦即進行調諧，可針對即使在製造工程也表面模樣不易變化之處重點地抽出特徵點。其結果，即使是實施在製造工程途中表面模樣變化的加工之類者，亦即在畫像的登錄時及識別時表面模樣變化的情況，也可取得能精度佳識別的良好效果。Since the individual recognition system 1 includes the visualization device 33, the feature points extracted by the feature point extraction process are displayed on the display device 12 by being superimposed on the image, so that the user can visually confirm where in the image the feature points are extracted. Then, the user can set various conditions or parameters while confirming the positions of the extracted feature points, that is, perform tuning, and can extract feature points with emphasis on areas where the surface pattern is not easily changed even in the manufacturing process. As a result, even in cases where the surface pattern changes during the manufacturing process, that is, the surface pattern changes during registration and recognition of an image, it is possible to obtain an excellent effect of being able to recognize with high accuracy.
藉由個體識別系統1具備攝影條件調整裝置31，使用者可操作攝影對象條件調整處理部312及攝影機器條件調整處理部311來調整適於特徵點的抽出及局部特徵量的計算之攝影對象條件及攝影機器條件。因此，藉由設定對應於環境的變化之適當的條件，可使干擾的影響減低，可使作為特徵的模樣適當地浮起。其結果，可取得能使識別精度提升這樣的良好效果。Since the individual recognition system 1 is provided with the photographing condition adjusting device 31, the user can operate the photographing target condition adjusting processing unit 312 and the photographing device condition adjusting processing unit 311 to adjust photographing target conditions suitable for extraction of feature points and calculation of local feature quantities. and camera conditions. Therefore, by setting appropriate conditions according to changes in the environment, the influence of disturbance can be reduced, and the characteristic pattern can be appropriately raised. As a result, it is possible to obtain a good effect that the recognition accuracy can be improved.
藉由個體識別系統1具備對照識別條件調整裝置32，使用者不須逐一手工作業作成參數的組合進行對照。而且，其對照結果是被彙整顯示於顯示裝置12。因此，使用者可藉由看被顯示於顯示裝置12的結果來確認各參數組的性能。然後，使用者是只要看被顯示於顯示裝置12的結果來選擇適當的參數組即可。藉此，可取得不論所欲對照識別的對象的種類，可簡單地不費工夫精度佳調整用在對照識別的算法的參數之良好的效果。Since the individual identification system 1 is provided with the comparison identification condition adjustment device 32, the user does not need to manually create a combination of parameters for comparison one by one. Furthermore, the comparison result is displayed on the display device 12 in a consolidated manner. Therefore, the user can confirm the performance of each parameter set by looking at the results displayed on the display device 12 . Then, the user only needs to select an appropriate parameter group by looking at the result displayed on the display device 12 . Thereby, regardless of the type of the object to be checked and identified, it is possible to easily and accurately adjust the parameters of the algorithm for matching and identifying without much effort.
而且，個體識別系統1是藉由同時具備可視化裝置33、攝影條件調整裝置31及對照識別條件調整裝置32的其中至少2個的裝置，可同時取得2個以上上述的各裝置的良好的效果。藉此，可對應於識別對象的表面模樣的變化，或不易受到干擾的影響，或減低花在條件或參數的調整之使用者工夫而可提高精度的條件或參數的精度，其結果，可提高對照識別的性能。Furthermore, the individual identification system 1 is a device that simultaneously includes at least two of the visualization device 33 , the imaging condition adjustment device 31 , and the comparison identification condition adjustment device 32 , so that two or more of the above-mentioned devices can simultaneously achieve good effects. Thereby, the accuracy of conditions or parameters can be improved in response to changes in the surface pattern of the recognition object, or it is less susceptible to interference, or the user's effort to adjust the conditions or parameters can be reduced, and the accuracy of the conditions or parameters can be improved. The performance of the control recognition.
個體識別系統1是具備可視化裝置33、攝影條件調整裝置31及對照識別條件調整裝置32。若根據此，則藉由全部同時具備可視化裝置33、攝影條件調整裝置31及對照識別條件調整裝置32，可全部同時取得上述的各裝置的良好的效果。亦即，若根據此個體識別系統1，則可對應於識別對象的表面形狀的變化，可進行不易受到干擾的影響之調整，進一步可減低花在條件或參數的調整之使用者工夫。The individual identification system 1 includes a visualization device 33 , an imaging condition adjustment device 31 , and a comparison identification condition adjustment device 32 . According to this, by simultaneously including the visualization device 33 , the imaging condition adjustment device 31 , and the collation recognition condition adjustment device 32 , all of the above-mentioned excellent effects of the devices can be obtained at the same time. That is, according to the individual identification system 1, it is possible to perform adjustment that is less susceptible to disturbances in response to changes in the surface shape of the object to be identified, and further reduces the user's effort in adjusting conditions and parameters.
檢索處理部261是更具有可實行鎖定處理的鎖定處理部2611。鎖定處理是從被登錄於檢索對象畫像登錄部241的複數的登錄畫像之中鎖定具有與識別畫像所具有的廣域特徵量的相關性高的廣域特徵量之預定數的登錄畫像作為特定處理候補之處理。The search processing unit 261 further includes a lock processing unit 2611 capable of executing lock processing. The locking process is to lock, as specific processing, a predetermined number of registered portraits having a wide-area feature that has a high correlation with the wide-area feature of the identification portrait from among the plurality of registered portraits registered in the search-target portrait registration unit 241 . Alternate processing.
若根據此，則檢索處理是可藉由實行：以廣域特徵量來鎖定的鎖定處理，及以局部特徵量來特定的特定處理之2階段的處理，實現高速的檢索。其結果，若根據本構成的個體識別系統1，則其結果，可實現對照識別性能高且高速的檢索。According to this, the retrieval processing can realize high-speed retrieval by executing two-stage processing of locking processing based on the wide-area feature value and specifying processing using the local feature value. As a result, according to the individual identification system 1 of the present configuration, as a result, high-speed retrieval with high matching identification performance can be realized.
本案是準照實施例敘述，但本案不是被限定於該實施例或構造者。本案還包含各種的變形例或均等範圍內的變形。加上，各種的組合或形態、進一步包含該等僅一要素或以上或以下的其他的組合或形態也進入本案的範疇或思想範圍。This case is described according to the embodiment, but the case is not limited to the embodiment or the constructor. The present case also includes various modifications or modifications within an equivalent range. In addition, various combinations or forms, and other combinations or forms that further include only one of the elements or more or less are also included in the scope or ideological scope of the present case.
本案記載的控制部及其手法是亦可藉由被程式化成實行藉由電腦程式來具體化的一個乃至複數的機能之構成處理器及記憶體而提供的專用電腦所實現。或，本案記載的控制部及其手法是亦可藉由一個以上的專用硬體邏輯電路來構成處理器而提供的專用電腦所實現。或，本案記載的控制部及其手法是亦可藉由被程式化成實行一個乃至複數的機能之處理器及記憶體與藉由一個以上的硬體邏輯電路來構成的處理器的組合而構成的一個以上的專用電腦所實現。又，電腦程式是亦可作為藉由電腦來實行的指令，被記憶於電腦可讀取的非遷移有形記錄媒體。The control unit and its method described in this case can also be realized by a dedicated computer that is programmed to implement one or more functions embodied by a computer program to constitute a processor and a memory provided. Alternatively, the control unit and its method described in this application may be realized by a dedicated computer provided by constituting a processor by one or more dedicated hardware logic circuits. Alternatively, the control unit and its method described in this case may be constituted by a combination of a processor and memory programmed to perform one or more functions, and a processor constituted by one or more hardware logic circuits Implemented by more than one dedicated computer. In addition, the computer program can also be memorized in a non-transportable tangible recording medium readable by a computer as an instruction to be executed by a computer.
