TWI801911B - Aquatic organism identification method and system - Google Patents
Aquatic organism identification method and system Download PDFInfo
- Publication number
- TWI801911B TWI801911B TW110122485A TW110122485A TWI801911B TW I801911 B TWI801911 B TW I801911B TW 110122485 A TW110122485 A TW 110122485A TW 110122485 A TW110122485 A TW 110122485A TW I801911 B TWI801911 B TW I801911B
- Authority
- TW
- Taiwan
- Prior art keywords
- image
- biometric
- biological
- underwater
- module
- Prior art date
Links
Images
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
- Breeding Of Plants And Reproduction By Means Of Culturing (AREA)
Abstract
Description
本發明係有關於一種生物辨識的技術領域,特別是有關於一種針對生物的各種外觀特徵來進行辨識的水中生物辨識方法及系統。 The present invention relates to the technical field of biological identification, in particular to an underwater biological identification method and system for identifying various appearance characteristics of living things.
生物辨識是目前經常用來對於生物群聚的環境中尋找特定的生物,並且對辨識出的生物進行追蹤及記錄生物長狀態。尤其是對於水中生物,水中生物的種類繁多,經常是生物研究的對象,因此生物辨識追蹤遂成為重要的研究課題。 Biometrics is currently often used to find specific organisms in the environment of biological clusters, and to track and record the growth status of the identified organisms. Especially for aquatic organisms, there are many kinds of aquatic organisms, which are often the objects of biological research, so biometric tracking has become an important research topic.
現有的生物辨識方法是對生物整體的影像進行辨識,因此在生物辨識系統進行深度學習時,需要利用大量圖片進行訓練,而且由於是針對生物整體進行辨識,因此生物的姿態也會影響辨識的結果。 The existing biometric identification method is to identify the image of the whole organism. Therefore, when the biometric system performs deep learning, it needs to use a large number of pictures for training, and because it is for the identification of the whole organism, the posture of the organism will also affect the identification result. .
有鑑於此,本發明提供一種水中生物辨識方法及系統,其藉由對生物的各種外觀特徵進行辨識,例如生物體的紋路特徵、顏色特徵及形態特徵,因此水中生物辨識系統只需要對生物體的各種外觀特徵進行深度訓練即可,不需要大量的圖片進行訓練。 In view of this, the present invention provides an underwater biometric identification method and system, which recognizes various appearance features of organisms, such as texture features, color features, and morphological features of organisms, so the underwater biometric identification system only needs to identify the biological Various appearance features of various appearance features can be trained in depth, and a large number of pictures are not required for training.
本發明的水中生物辨識方法的一實施例包括:連續擷取多個第一影像;判斷該等第一影像中是否具有水中生物影像;若具有水中生物影像,則擷取該水中生物影像的一生物特徵影像;將所擷取的該生物特徵影像與一標準生物特徵影像比對,該標準生物特徵影像係對應於一水中生物;判斷該生物特徵影像是否近似於該標準生物特徵影像;以及若該生物特徵近似於該標準生物特徵影像,則顯示該水中生物。 An embodiment of the aquatic organism identification method of the present invention includes: continuously capturing a plurality of first images; judging whether there are aquatic organism images in the first images; if there are aquatic organism images, capturing one of the aquatic organism images a biometric image; comparing the captured biometric image with a standard biometric image corresponding to an aquatic organism; judging whether the biometric image is similar to the standard biometric image; and if If the biological feature is similar to the standard biological feature image, the aquatic organism is displayed.
在另一實施例中,該生物特徵影像包括紋路特徵影像、顏色特徵影像及/或形態特徵影像。 In another embodiment, the biological feature image includes texture feature image, color feature image and/or morphological feature image.
在另一實施例中,擷取該水中生物影像的物種特徵影像;將該物種特徵影像與一標準物種特徵影像比對;若該物種特徵影像與該標準物種特徵影像近似,則擷取該水中生物影像的該生物特徵影像。 In another embodiment, the species feature image of the underwater biological image is captured; the species feature image is compared with a standard species feature image; if the species feature image is similar to the standard species feature image, the underwater The biometric image of the biometric image.
在另一實施例中,該標準生物特徵影像及該標準物種特徵影像係以多數個該水中生物的標準影像獲得。 In another embodiment, the standard biological feature image and the standard species feature image are obtained from a plurality of standard images of the aquatic organisms.
在另一實施例中,判斷該等第一影像中是否具有該水中生物影像的步驟係判斷該等第一影像中的特定區域是否具有該水中生物影像。 In another embodiment, the step of judging whether the image of the aquatic organism exists in the first images is to determine whether the image of the aquatic organism exists in a specific area of the first images.
在另一實施例中,動態追蹤該水中生物一既定時間;判斷該水中生物的尺寸是否發生變化;判斷該水中生物在該既定時間內的移動量。 In another embodiment, the aquatic organism is dynamically tracked for a predetermined time; whether the size of the aquatic organism changes; and the movement of the aquatic organism within the predetermined time is determined.
在另一實施例中,該紋路特徵影像係根據局部二值化模式(Local Binary Pattern,LBP)做為其計算的基礎理論。 In another embodiment, the texture feature image is based on a Local Binary Pattern (LBP) as the basic theory for its calculation.
在另一實施例中,該紋路特徵影像係根據灰階共生矩陣(Gray Level Co-occurrence Matrix,GLCM)做為其計算的基礎理論。 In another embodiment, the texture feature image is calculated based on a Gray Level Co-occurrence Matrix (GLCM).
在另一實施例中,該紋路特徵影像係根據局部模式共生矩陣(Local Pattern Co-occurrence Matrix,LPCM)做為其計算的基礎理論。 In another embodiment, the texture feature image is calculated based on a Local Pattern Co-occurrence Matrix (LPCM).
在另一實施例中,該紋路特徵影像係根據紋理基元法(Texton-Based Approach,TBA)做為其計算的基礎理論。 In another embodiment, the texture feature image is based on the Texton-Based Approach (TBA) as the basic theory for its calculation.
本發明的水中生物辨識系統包括:一影像擷取模組、一生物辨識模組、一特徵擷取模組、一比對模組以及一影像輸出模組。影像擷取模組對一區域連續擷取多個第一影像。生物辨識模組對該等第一影像進行辨識,以判斷該等第一影像中是否具有水中生物影像,並且將具有水中生物影像的第一影像作為第二影像。特徵擷取模組擷取該第二影像的該水中生物影像的生物特徵影像。比對模組將所擷取的該生物特徵影像與一標準生物特徵影像比對,並判斷該生物特徵影像是否近似於該標準生物特徵影像,該標準生物特徵影像對應於一水中生物。影像輸出模組,當該生物特徵影像近似於該標準生物特徵影像時,顯示該第一影像的該水中生物。 The underwater biological identification system of the present invention includes: an image capture module, a biometric identification module, a feature extraction module, a comparison module and an image output module. The image capturing module continuously captures a plurality of first images for an area. The biometric identification module recognizes the first images to determine whether there are images of aquatic organisms in the first images, and uses the first image with images of aquatic organisms as the second image. The feature extraction module extracts the biometric image of the underwater biological image of the second image. The comparison module compares the captured biometric image with a standard biometric image, and judges whether the biometric image is similar to the standard biometric image, and the standard biometric image corresponds to an aquatic organism. The image output module displays the aquatic organisms in the first image when the biometric image is similar to the standard biometric image.
在另一實施例中,本發明的水中生物辨識系統更包括一區域限定模組,其限定該生物辨識模組針對該等第一影像中的一特定區域進行辨識。 In another embodiment, the underwater biological identification system of the present invention further includes an area limiting module, which limits the biological identification module to identify a specific area in the first images.
在另一實施例中,本發明的水中生物辨識系統更包括一動態追蹤模組,根據比對模組比對的結果動態追蹤該水中生物一既定時間,並判斷水中生物的尺寸是否發生變化以及計算該水中生物在該既定時間內的移動量。 In another embodiment, the underwater biological identification system of the present invention further includes a dynamic tracking module, which dynamically tracks the aquatic organisms for a predetermined period of time according to the comparison result of the comparison module, and determines whether the size of the aquatic organisms changes and Calculate the amount of movement of the aquatic organism within the given time.
在另一實施例中,本發明的水中生物辨識系統更包括一生物辨識訓練資料庫、一特徵擷取訓練資料庫以及一比對訓練資料庫。生物辨識訓練資料庫儲存有複數個生物影像資料,該生物辨識模組根據該等生物影像資料進行辨識的學習。特徵擷取訓練資料庫儲存有複數個生物特徵影像資料,該特徵擷取模 組根據該等生物特徵影像資料進行影像擷取的學習。比對訓練資料庫儲存有複數個生物特徵影像資料,該比對模組根據該等生物特徵影像資料進行比對的學習。 In another embodiment, the underwater biometric identification system of the present invention further includes a biometric identification training database, a feature extraction training database, and a comparison training database. The biometric identification training database stores a plurality of biological image data, and the biometric identification module performs identification learning according to the biological image data. The feature extraction training database stores a plurality of biometric image data, and the feature extraction model The group conducts image acquisition learning based on the biometric image data. The comparison training database stores a plurality of biometric image data, and the comparison module performs comparison learning according to the biometric image data.
在另一實施例中,更包含一拼接模組,係對複數個該第二影像進行拼接,以形成具有完整態樣的生物影像之該第二影像。 In another embodiment, it further includes a splicing module for splicing a plurality of the second images to form the second image with a complete biological image.
本發明的水中生物辨識方法及水中生物辨識系統係藉由對生物體的特徵進行辨識,尤其是針對外觀特徵進行辨識,例如生物體的紋路特徵、顏色特徵及形態特徵,因此水中生物辨識系統只需要對生物體的各種外觀特徵進行深度訓練即可,不需要大量的圖片進行訓練,不僅增加辨識的準確率,而且可以避免因為生物體在活動時產生不同的姿態而影響辨識的結果。 The underwater biometrics identification method and underwater biometrics system of the present invention identify the characteristics of organisms, especially for the identification of appearance features, such as texture features, color features and morphological features of organisms, so the underwater biometrics system only It is only necessary to conduct in-depth training on various appearance features of organisms, and does not require a large number of pictures for training, which not only increases the accuracy of recognition, but also avoids affecting the recognition results due to different postures of organisms when they move.
10:影像擷取模組 10: Image capture module
20:生物辨識模組 20: Biometric module
21:生物辨識訓練資料庫 21: Biometric training database
30:特徵擷取模組 30: Feature extraction module
31:特徵擷取訓練資料庫 31: Feature extraction training database
40:比對模組 40: Compare modules
41:比對訓練資料庫 41: Comparing the training database
50:影像輸出模組 50: Image output module
60:處理器 60: Processor
70:無線通訊模組 70: Wireless communication module
80:儲存模組 80:Storage module
90:雲端資料庫 90:Cloud database
100:水中生物辨識系統 100: Water biometrics system
110:區域限定模組 110: Region-limited modules
120:時間記錄模組 120: Time recording module
130:動態追蹤模組 130:Dynamic tracking module
A1、A2、A3、B1、B2、C1、C2:水中生物 A1, A2, A3, B1, B2, C1, C2: aquatic organisms
F1~F7:水中生物 F1~F7: Aquatic creatures
FA:長大的水中生物 FA: grown-up aquatic organisms
FB:年幼的水中生物 FB: young aquatic creatures
S11~S15、S21~S27、S31~S36:步驟 S11~S15, S21~S27, S31~S36: steps
D:框體 D: frame
第1圖為本發明的水中生物辨識方法的第一實施例的流程圖。 Fig. 1 is a flow chart of the first embodiment of the underwater biometric identification method of the present invention.
第2圖為本發明的水中生物辨識方法的第二實施例的流程圖。 Fig. 2 is a flow chart of the second embodiment of the underwater biometric identification method of the present invention.
第3圖為本發明的水中生物辨識方法的第三實施例的流程圖。 Fig. 3 is a flow chart of the third embodiment of the underwater biometric identification method of the present invention.
第4圖為本發明的水中生物辨識方法的紋路特徵辨識的運算基礎的示意圖。 Fig. 4 is a schematic diagram of the calculation basis of texture feature recognition in the underwater biological recognition method of the present invention.
第5圖為本發明的水中生物辨識系統的第一實施例的方塊圖。 Fig. 5 is a block diagram of the first embodiment of the underwater biometric identification system of the present invention.
第6圖為本發明的水中生物辨識系統的第二實施例的方塊圖。 Fig. 6 is a block diagram of the second embodiment of the underwater biometric identification system of the present invention.
第7圖為本發明的水中生物辨識系統的第三實施例的方塊圖。 Fig. 7 is a block diagram of the third embodiment of the underwater biometric identification system of the present invention.
第8圖表示擷取一區域的影像,並對影像進行辨識。 FIG. 8 shows capturing an image of an area and recognizing the image.
第9圖表示該區域的影像經過辨識後,該影像具有水中生物的影像。 Fig. 9 shows that after the image of the area is recognized, the image has images of aquatic organisms.
第10圖表示該水中生物影像以生物特徵進行辨識後,顯示該水中生物的名稱。 Fig. 10 shows that after the image of the aquatic organism is identified by the biological characteristics, the name of the aquatic organism is displayed.
第11圖表示對水中生物進行辨識後,對希望追蹤的水中生物進行追蹤,並獲得水中生物的健康狀態。 Figure 11 shows that after identifying the aquatic organisms, the desired aquatic organisms are tracked, and the health status of the aquatic organisms is obtained.
第12圖表示各種不同水中生物經過辨識後分別標示顯示的狀態。 Fig. 12 shows the status of marking and displaying of various aquatic organisms after identification.
第13圖表示追蹤水中生物經過一段時間,觀察水中生物從幼魚到成魚。 Fig. 13 shows tracking aquatic organisms through a period of time, observing aquatic organisms from juveniles to adults.
請參閱第1圖,其表示本發明的水中生物辨識方法的第一實施例。本發明的生物辨識方法係利用已知的各種生物特徵對影像中的水中生物進行辨識。在步驟S11中,影像擷取模組擷取多張影像,接著進入步驟S12,在步驟S12中,請一併參閱第8圖,生物辨識模組篩選多張影像,辨識是否有水中生物。如果辨識結果是具有水中生物,則進入步驟S13,如果沒有辨識到水中生物,則捨棄該影像。在步驟S13中,請一併參閱第9圖,先對已辨識出的水中生物進行標示,例如框選水中生物F1、F2、F3,然後特徵擷取模組針對具有水中生物的影像進行擷取生物特徵,例如紋路特徵、顏色特徵或/以及形態特徵,紋路特徵為水中生物身上的花紋,顏色特徵為水中生物體的顏色,形態特徵為水中生物體的形狀。接著進入步驟S14,在步驟S14中,比對模組將擷取的特徵與比對訓練資料中的已知的生物特徵(紋路特徵、顏色特徵或/以及形態特徵)進行比對,如果辨識到與已知的生物特徵相符的水中生物,則進入步驟S15,如果未辨識到與已知的生物特徵相符的水中生物,則捨棄該影像,若該影像係具有完整態樣的水中
生物體時,該影像可儲存於儲存模組,以供下一次特徵擷取模組針對具有水中生物的影像進行擷取有別於該次的生物特徵;若該影像僅係具有部分態樣的水中生物體時,例如每一張影像分別為水中生物的頭部、尾部或身體,可藉由一拼接模組(圖未繪出),將部分態樣的水中生物體的影像進行拼接以形成一完整態樣的水中生物體的影像,並儲存於儲存模組,以供下一次特徵擷取模組針對具有水中生物的影像進行擷取有別於該次的生物特徵。在步驟S15中,請一併參閱第10圖,影像輸出模組將比對結果輸出並顯示其身分,在框體D的外部顯示水中生物F1、F2、F3的名稱,例如小丑魚1、小丑魚2和小丑魚3。
Please refer to Figure 1, which shows the first embodiment of the underwater biometric identification method of the present invention. The biological identification method of the present invention utilizes various known biological features to identify aquatic organisms in images. In step S11, the image capture module captures multiple images, and then enters step S12. In step S12, please refer to FIG. 8, the biometric identification module screens multiple images to identify whether there are aquatic organisms. If the identification result is that there are aquatic organisms, then go to step S13, and if no aquatic organisms are identified, discard the image. In step S13, please also refer to Figure 9, first mark the identified aquatic organisms, for example, select the aquatic organisms F1, F2, F3, and then the feature extraction module extracts images with aquatic organisms Biological characteristics, such as texture characteristics, color characteristics or/and morphological characteristics, the texture characteristics are the patterns on the aquatic organisms, the color characteristics are the colors of the aquatic organisms, and the morphological characteristics are the shapes of the aquatic organisms. Then enter step S14. In step S14, the comparison module compares the extracted features with the known biological features (texture features, color features or/and morphological features) in the comparison training data. For aquatic organisms that match the known biological characteristics, enter step S15. If no aquatic organisms that match the known biological characteristics are identified, discard the image. If the image is an underwater organism with a complete appearance
In the case of living organisms, the image can be stored in the storage module for the next feature extraction module to capture the biological characteristics different from the current image for the image with aquatic organisms; if the image only has part of the shape In the case of aquatic organisms, for example, each image is the head, tail or body of the aquatic organism, and a splicing module (not shown in the figure) can be used to splice the images of some forms of aquatic organisms to form An image of a complete aquatic organism is stored in the storage module for the next feature extraction module to extract biological features different from the image with aquatic organisms. In step S15, please also refer to Figure 10, the image output module will output the comparison result and display its identity, and display the names of aquatic organisms F1, F2, F3 outside the frame D, such as
請參閱第2圖,其表示本發明的水中生物辨識方法的第二實施例。在步驟S21中,影像擷取模組擷取多張影像,接著進入步驟S22,在步驟S22中,請一併參閱第8圖,生物辨識模組,篩選多張影像,辨識是否有水中生物。如果辨識結果是具有水中生物F1、F2、F3,則進入步驟S23,如果沒有辨識到水中生物,則捨棄該影像,若該影像係具有完整態樣的水中生物體時,該影像可儲存於儲存模組,以供下一次特徵擷取模組針對具有水中生物的影像進行擷取有別於該次的生物特徵;若該影像僅係具有部分態樣的水中生物體時,例如每一張影像分別為水中生物的頭部、尾部或身體,可藉由一拼接模組(圖未繪出),將部分態樣的水中生物體的影像進行拼接以形成一完整態樣的水中生物體的影像,並儲存於儲存模組,以供下一次特徵擷取模組針對具有水中生物的影像進行擷取有別於該次的生物特徵。在步驟S23中,特徵擷取模組針對具有水中生物的影像進行擷取個別物種特徵,接著進入步驟S24,在步驟S24中,比對模組將擷取的物種特徵與比對訓練資料中的已知的個別物種特徵進行比對,如果擷取的物種特徵與已知的個別物種特徵相符,則進入步驟S25,如果不相符,則捨棄該影像,
若該影像係具有完整態樣的水中生物體時,該影像可儲存於儲存模組,以供下一次特徵擷取模組針對具有水中生物的影像進行擷取有別於該次的生物特徵;上述的物種特徵係指在該區域中有多種生物體,例如在水族箱中有魚類、貝類、海星等不同的生物物種,例如目前希望辨識的是魚類的物種,則上述物種特徵則可以是魚鰭、魚尾等物種特徵。在步驟S25中,請一併參閱第9圖,先對已辨識出的水中生物進行標示,例如框選水中生物F1、F2、F3,然後特徵擷取模組針對具有水中生物的影像進行擷取生物特徵,例如紋路特徵、顏色特徵或/以及形態特徵,紋路特徵為水中生物身上的花紋,顏色特徵為水中生物體的顏色,形態特徵為水中生物體的形狀。接著進入步驟S26,在步驟S26中,比對模組將擷取的特徵與比對訓練資料中的已知的生物特徵(紋路特徵、顏色特徵或/以及形態特徵)進行比對,如果辨識到與已知的生物特徵相符的水中生物,則進入步驟S27,如果未辨識到與已知的生物特徵相符的水中生物,則捨棄該影像。在步驟S27中,請一併參閱第10圖,影像輸出模組將比對結果輸出並顯示其身分,在框體D的外部顯示水中生物F1、F2、F3的名稱,例如小丑魚1、小丑魚2和小丑魚3。
Please refer to FIG. 2, which shows the second embodiment of the underwater biometric identification method of the present invention. In step S21, the image capture module captures multiple images, and then enters step S22. In step S22, please refer to FIG. 8, the biometric module screens multiple images to identify whether there are aquatic organisms. If the identification result is that there are aquatic organisms F1, F2, and F3, then enter step S23. If no aquatic organisms are identified, the image is discarded. If the image is an aquatic organism with a complete appearance, the image can be stored in the storage. module for the next feature extraction module to extract biological features different from the image with aquatic organisms; if the image is only part of the aquatic organisms, for example, each image They are the head, tail or body of the aquatic organisms respectively. A splicing module (not shown in the figure) can be used to splice the images of the aquatic organisms in part to form a complete image of the aquatic organisms , and stored in the storage module, for the next feature extraction module to extract biological features different from this time for the image with aquatic organisms. In step S23, the feature extraction module extracts the characteristics of individual species from the image with aquatic organisms, and then enters step S24. In step S24, the comparison module compares the extracted species features with those in the training data. The known individual species characteristics are compared, if the extracted species characteristics are consistent with the known individual species characteristics, then enter step S25, if not, discard the image,
If the image has a complete form of aquatic organisms, the image can be stored in the storage module for the next feature extraction module to extract biological features different from the image with aquatic organisms; The above-mentioned species characteristics mean that there are many kinds of organisms in the area, for example, there are different biological species such as fish, shellfish, and starfish in the aquarium. Species characteristics such as fins and tails. In step S25, please also refer to Figure 9, first mark the identified aquatic organisms, for example, select the aquatic organisms F1, F2, and F3, and then the feature extraction module extracts images with aquatic organisms Biological characteristics, such as texture characteristics, color characteristics or/and morphological characteristics, the texture characteristics are the patterns on the aquatic organisms, the color characteristics are the colors of the aquatic organisms, and the morphological characteristics are the shapes of the aquatic organisms. Then enter step S26. In step S26, the comparison module compares the extracted features with the known biological features (texture features, color features or/and morphological features) in the comparison training data. For aquatic organisms that match the known biological features, proceed to step S27. If no aquatic organisms that match the known biological features are identified, the image is discarded. In step S27, please also refer to Figure 10, the image output module will output the comparison result and display its identity, and display the names of aquatic organisms F1, F2, F3 outside the frame D, such as
請參與第3圖,其表示本發明的水中生物辨識方法的第三實施例。本發明的生物辨識方法係利用已知的各種生物特徵對影像中的水中生物進行辨識。在步驟S31中,影像擷取模組擷取多張影像,接著進入步驟S32。在步驟S32中,區域限定模組限定該影像的特定區域,例如將影像限定在水族箱的區域,接著進入步驟S33。在步驟S33中,請一併參閱第8圖,生物辨識模組篩選多張影像,辨識是否有水中生物。如果辨識結果是具有水中生物,則進入步驟S34,如果沒有辨識到水中生物,則捨棄該影像。在步驟S34中,請一併參閱第9圖,先對已辨識出的水中生物F1、F2、F3進行標示,例如框選水中生物,然後特
徵擷取模組針對具有水中生物的影像進行擷取生物特徵,例如紋路特徵、顏色特徵或/以及形態特徵,紋路特徵為水中生物身上的花紋,顏色特徵為水中生物體的顏色,形態特徵為水中生物體的形狀。接著進入步驟S35,在步驟S35中,比對模組將擷取的特徵與比對訓練資料中的已知的生物特徵(紋路特徵、顏色特徵或/以及形態特徵)進行比對,如果辨識到與已知的生物特徵相符的水中生物,則進入步驟S36,如果未辨識到與已知的生物特徵相符的水中生物,則捨棄該影像,若該影像係具有完整態樣的水中生物體時,該影像可儲存於儲存模組,以供下一次特徵擷取模組針對具有水中生物的影像進行擷取有別於該次的生物特徵;若該影像僅係具有部分態樣的水中生物體時,例如每一張影像分別為水中生物的頭部、尾部或身體,可藉由一拼接模組(圖未繪出),將部分態樣的水中生物體的影像進行拼接以形成一完整態樣的水中生物體的影像,並儲存於儲存模組,以供下一次特徵擷取模組針對具有水中生物的影像進行擷取有別於該次的生物特徵。在步驟S36中,請一併參閱第10圖,影像輸出模組將比對結果輸出並顯示其身分,在框體D的外部顯示水中生物F1、F2、F3的名稱,例如小丑魚1、小丑魚2和小丑魚3。
Please refer to Figure 3, which represents the third embodiment of the underwater biometric identification method of the present invention. The biological identification method of the present invention utilizes various known biological features to identify aquatic organisms in images. In step S31, the image capture module captures a plurality of images, and then enters step S32. In step S32 , the region limiting module defines a specific region of the image, for example, the image is restricted to the area of the aquarium, and then proceeds to step S33 . In step S33, please also refer to FIG. 8 , the biometric identification module screens multiple images to identify whether there are aquatic organisms. If the identification result is that there are aquatic organisms, then go to step S34, and if no aquatic organisms are identified, discard the image. In step S34, please also refer to Fig. 9, first mark the identified aquatic organisms F1, F2, F3, for example select the aquatic organisms, and then specifically
The levy extraction module extracts biological features from images with aquatic organisms, such as texture features, color features, and/or morphological features. The texture features are patterns on aquatic organisms, the color feature is the color of aquatic organisms, and the morphological features are The shape of organisms in water. Then enter step S35. In step S35, the comparison module compares the extracted features with the known biological features (texture features, color features or/and morphological features) in the comparison training data. For aquatic organisms that match the known biological characteristics, enter step S36. If no aquatic organisms that match the known biological characteristics are identified, the image is discarded. If the image is an aquatic organism with a complete appearance, The image can be stored in the storage module for the next feature extraction module to extract biological features different from the image with aquatic organisms; if the image is only partially aquatic organisms , for example, each image is the head, tail or body of an aquatic organism, and a splicing module (not shown in the figure) can be used to splice the images of some aquatic organisms to form a complete image The images of the aquatic organisms are stored in the storage module for the next feature extraction module to extract biological features different from the images of the aquatic organisms. In step S36, please also refer to Figure 10, the image output module will output the comparison result and display its identity, and display the names of aquatic organisms F1, F2, F3 outside the frame D, such as
請參閱第4圖,紋路特徵可以根據局部二值化模式(Local Binary Pattern,LBP)、灰階共生矩陣(Gray Level Co-occurrence Matrix,GLCM)、局部模式共生矩陣(Local Pattern Co-occurrence Matrix,LPCM)以及紋理基元法(Texton-Based Approach,TBA)做為其計算的基礎理論。局部二值化模式是一種描述材質局部紋理的特徵抽取方式,將像素與鄰近像素的相互關係利用二值化進行編碼,其具有光亮及旋轉不變。灰階共生矩陣是統計像素在空間中相對位置的灰階值變化,反映紋理在空間中分布。局部模式共生矩陣原理是將局部紋理特性空間灰 階關係做結合,藉此提高特徵抽取解決環境變化問題的能力。紋理基元法的原理是找出材質圖片中的紋理基元,統計其個數並得到紋理特徵。 Please refer to Figure 4. The texture features can be based on the Local Binary Pattern (LBP), Gray Level Co-occurrence Matrix (GLCM), Local Pattern Co-occurrence Matrix (Local Pattern Co-occurrence Matrix, LPCM) and Texton-Based Approach (TBA) as the basic theory of its calculation. The local binarization mode is a feature extraction method that describes the local texture of the material. The relationship between pixels and adjacent pixels is encoded by binarization, which is invariant to brightness and rotation. The gray-scale co-occurrence matrix is the gray-scale value change of the relative position of the statistical pixel in space, reflecting the distribution of texture in space. The principle of the local mode co-occurrence matrix is to gray out the local texture characteristic space The order relationship is combined to improve the ability of feature extraction to solve environmental changes. The principle of the texture primitive method is to find out the texture primitives in the material picture, count their number and obtain the texture features.
請參閱第5圖,其表示本發明的水中生物辨識系統的第一實施例。本發明的水中生物辨識系統100包括:一影像擷取模組10、一生物辨識模組20、一特徵擷取模組30、一比對模組40以及一影像輸出模組50。影像擷取模組10對一區域連續擷取多個第一影像,影像擷取模組10可以是任何攝影裝置,如第11圖所示。第一影像被傳送至生物辨識模組20對該等第一影像進行辨識,以判斷該等第一影像中是否具有水中生物影像,並且將具有水中生物影像的第一影像作為第二影像。特徵擷取模組30擷取該第二影像的該水中生物影像的生物特徵影像。比對模組40將所擷取的該生物特徵影像與一標準生物特徵影像比對,並判斷該生物特徵影像是否近似於該標準生物特徵影像,該標準生物特徵影像對應於一水中生物。當該生物特徵影像近似於該標準生物特徵影像時,影像輸出模組50顯示該第一影像的水中生物,影像輸出模組50可以是一顯示裝置。本實施例的水中生物辨識系統100還包括一處理器60,生物辨識模組20、特徵擷取模組30及比對模組40可以是程式模組,這些程式模組可以在處理器60中執行。水中生物辨識系統100還包括一生物辨識訓練資料庫21、一特徵擷取訓練資料庫31以及一比對訓練資料庫41。生物辨識訓練資料庫21儲存有複數個生物影像資料,該生物辨識模組20根據該等生物影像資料進行辨識的學習。特徵擷取訓練資料庫31儲存有複數個生物特徵影像資料,該特徵擷取模組30根據該等生物特徵影像資料進行影像擷取的學習。比對訓練資料庫41儲存有複數個生物特徵影像資料,比對模組40根據該等生物特徵影像資料進行比對的學習,上述所提辨識的學習、擷取的學習或比對的學習可舉例但不限定於使用支援向量機(support vector machine,SVM)。
Please refer to Fig. 5, which shows the first embodiment of the underwater biometric identification system of the present invention. The underwater
本實施例的水中生物辨識系統100還包括無線通訊模組70和儲存模組80,儲存模組80可以儲存所擷取的影像擷取模組10擷取的影像資料。無線通訊模組70訊號連接於一雲端資料庫90。水中生物辨識系統100經由無線通訊模組70連接於雲端資料庫90,辨識後的影像資料可以傳送至雲端資料庫90進行分析,本創作一實施例中,水中生物辨識系統100也可以從雲端資料庫90下載新的生物辨識訓練資料、新的特徵擷取訓練資料和新的比對訓練資料,先儲存於儲存模組80,然後再分別儲存至生物辨識訓練資料庫21、一特徵擷取訓練資料庫31以及一比對訓練資料庫41,分別供生物辨識模組20、特徵擷取模組30及比對模組40進行學習。如第12圖所示,對不同種類的水中生物可以分別標示,例如水中生物A1、A2、A3、B1、B2、C1、C2。
The underwater
請參閱第6圖,其表示本發明的水中生物辨識系統的第二實施例。本實施例的水中生物辨識系統與第5圖的水中生物辨識系統具有部分相同的結構,因此相同的元件給予相同的標號,並省略其說明。本實施例與第一實施例的差異在於本實施例更包括一區域限定模組110,其限定該生物辨識模組20針對該等第一影像中的一特定區域進行辨識。如上所述,例如對水族箱區域進行辨識。
Please refer to Figure 6, which shows the second embodiment of the underwater biometric identification system of the present invention. The underwater biometric identification system of this embodiment has part of the same structure as the aquatic biometric identification system of FIG. 5 , so the same components are given the same reference numerals, and their descriptions are omitted. The difference between this embodiment and the first embodiment is that this embodiment further includes an
請參閱第7圖,其表示本發明的水中生物辨識系統的第三實施例。本實施例的水中生物辨識系統與第5圖的水中生物辨識系統具有部分相同的結構,因此相同的元件給予相同的標號,並省略其說明。本實施例與第一實施例的差異在於本實施例更包括時間記錄模組120和動態追蹤模組130。時間記錄模組120可以記錄追蹤的時間,動態追蹤模組130根據比對模組比對的結果動態追蹤該水中生物一既定時間,並判斷水中生物F4、F5、F6、F7的尺寸是否發生變化以及計算該水中生物在該既定時間內的移動量。如第11圖所示,在一段時間內,如果
水中生物F5、F7移動量變低,則判定該水中生物生病,也可以顯示於影像輸出模組50。如第13圖所示,在一段時間內,年幼的水中生物FB的尺寸發生變化變成長大的水中生物FA,即可判定水中生物成長。
Please refer to Fig. 7, which shows the third embodiment of the underwater biometric identification system of the present invention. The underwater biometric identification system of this embodiment has part of the same structure as the aquatic biometric identification system of FIG. 5 , so the same components are given the same reference numerals, and their descriptions are omitted. The difference between this embodiment and the first embodiment is that this embodiment further includes a
本發明的水中生物辨識方法及水中生物辨識系統係藉由對生物體的特徵進行辨識,尤其是針對外觀特徵進行辨識,例如生物體的紋路特徵、顏色特徵及形態特徵,因此水中生物辨識系統只需要對生物體的各種外觀特徵進行深度訓練即可,不需要大量的圖片進行訓練,不僅增加辨識的準確率,而且可以避免因為生物體在活動時產生不同的姿態而影響辨識的結果。 The underwater biometrics identification method and underwater biometrics system of the present invention identify the characteristics of organisms, especially for the identification of appearance features, such as texture features, color features and morphological features of organisms, so the underwater biometrics system only It is only necessary to conduct in-depth training on various appearance features of organisms, and does not require a large number of pictures for training, which not only increases the accuracy of recognition, but also avoids affecting the recognition results due to different postures of organisms when they move.
S11~S15:步驟 S11~S15: Steps
Claims (12)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW110122485A TWI801911B (en) | 2021-06-18 | 2021-06-18 | Aquatic organism identification method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW110122485A TWI801911B (en) | 2021-06-18 | 2021-06-18 | Aquatic organism identification method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
TW202301184A TW202301184A (en) | 2023-01-01 |
TWI801911B true TWI801911B (en) | 2023-05-11 |
Family
ID=86658154
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
TW110122485A TWI801911B (en) | 2021-06-18 | 2021-06-18 | Aquatic organism identification method and system |
Country Status (1)
Country | Link |
---|---|
TW (1) | TWI801911B (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2017221259A1 (en) * | 2016-06-23 | 2017-12-28 | S Jyothi | Automatic recognition of indian prawn species |
CN111753775A (en) * | 2020-06-29 | 2020-10-09 | 北京海益同展信息科技有限公司 | Fish growth assessment method, device, equipment and storage medium |
CN111968159A (en) * | 2020-08-28 | 2020-11-20 | 厦门大学 | Simple and universal fish video image track tracking method |
CN112598713A (en) * | 2021-03-03 | 2021-04-02 | 浙江大学 | Offshore submarine fish detection and tracking statistical method based on deep learning |
-
2021
- 2021-06-18 TW TW110122485A patent/TWI801911B/en active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2017221259A1 (en) * | 2016-06-23 | 2017-12-28 | S Jyothi | Automatic recognition of indian prawn species |
CN111753775A (en) * | 2020-06-29 | 2020-10-09 | 北京海益同展信息科技有限公司 | Fish growth assessment method, device, equipment and storage medium |
CN111968159A (en) * | 2020-08-28 | 2020-11-20 | 厦门大学 | Simple and universal fish video image track tracking method |
CN112598713A (en) * | 2021-03-03 | 2021-04-02 | 浙江大学 | Offshore submarine fish detection and tracking statistical method based on deep learning |
Also Published As
Publication number | Publication date |
---|---|
TW202301184A (en) | 2023-01-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Sun et al. | Transferring deep knowledge for object recognition in low-quality underwater videos | |
US6421463B1 (en) | Trainable system to search for objects in images | |
CN108596046A (en) | A kind of cell detection method of counting and system based on deep learning | |
Yu et al. | An object-based visual attention model for robotic applications | |
WO2019245722A1 (en) | Sea lice detection and classification in an aquaculture environment | |
CN108154102A (en) | A kind of traffic sign recognition method | |
CN109410168A (en) | For determining the modeling method of the convolutional neural networks model of the classification of the subgraph block in image | |
US20230071265A1 (en) | Quantifying plant infestation by estimating the number of biological objects on leaves, by convolutional neural networks that use training images obtained by a semi-supervised approach | |
CN105825168B (en) | A kind of Rhinopithecus roxellana face detection and method for tracing based on S-TLD | |
Qiao et al. | Bird species recognition based on SVM classifier and decision tree | |
Li et al. | CMFTNet: Multiple fish tracking based on counterpoised JointNet | |
CN115578423A (en) | Fish key point detection, individual tracking and biomass estimation method and system based on deep learning | |
Sridharan et al. | Real-time vision on a mobile robot platform | |
CN112528823B (en) | Method and system for analyzing batcharybus movement behavior based on key frame detection and semantic component segmentation | |
Liu et al. | Research progress of computer vision technology in abnormal fish detection | |
Taylor et al. | Pose-sensitive embedding by nonlinear nca regression | |
TWI801911B (en) | Aquatic organism identification method and system | |
Mathisen et al. | FishNet: A unified embedding for Salmon recognition | |
CN111738062A (en) | Automatic re-identification method and system based on embedded platform | |
CN108573226B (en) | Drosophila larva body node key point positioning method based on cascade posture regression | |
CN113673422A (en) | Pet type identification method and identification system | |
CN109063591A (en) | A kind of recognition methods again of the pedestrian based on range distribution metric learning | |
US20230230258A1 (en) | Materials and methods for long-term tracking of group-housed livestock | |
Prow | Seeing the Trees from the Forest: Using Modern Methods to Identify Individual Objects in a Cluttered Environment for Robots | |
Braun | Tracking multiple mice |