CN101276404A - System and method for quickly and exactly processing intelligent image - Google Patents
System and method for quickly and exactly processing intelligent image Download PDFInfo
- Publication number
- CN101276404A CN101276404A CNA2007100388464A CN200710038846A CN101276404A CN 101276404 A CN101276404 A CN 101276404A CN A2007100388464 A CNA2007100388464 A CN A2007100388464A CN 200710038846 A CN200710038846 A CN 200710038846A CN 101276404 A CN101276404 A CN 101276404A
- Authority
- CN
- China
- Prior art keywords
- image
- matrix
- vector
- module
- intelligent
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Image Processing (AREA)
Abstract
The invention relates to a rapidly and accurately intelligent image processing system, comprising a plurality of modules of image capture, image converting, data transmission, calculation, storage processing, data transmission and data output; the processing method thereof comprises: step 1, if the analysis model has been loaded, executing step 4, or executing step 2; step 2, obtaining registered image and properties thereof, and carrying out target detection, location and normalizing preprocess to obtain a sample characteristic matrix; step 3, saving and loading system according to the analysis model; step 4, receiving image to be identified and carrying out target detection, location, normalizing preprocess and characteristic extraction; step 5, calculating a judging vector according to the model; step 6, using index corresponding to the maximal element in the vector as property number of the target image; the system and processing method is characterized of low storage requirement, rapid identifying speed, high identifying accuracy, suitable for industrial application or public utilities field with obvious economic benefits.
Description
Technical field
The present invention relates to a kind of disposal system of intelligent image fast and accurately and disposal route thereof, belong to Digital Image Processing, data analysis technique field.
Background technology
The intelligent image treatment technology is widely used in the computer vision association area, such as people's face identification based on image, human face expression identification, face gender identification and people's face estimation of Age etc., wherein the identification of people's face can be used for the certificate photograph affirmation, building turnover control, fields such as credit card identity validation, human face expression identification, face gender identification, people's face estimation of Age can be used for various intelligent human-machine interaction system, and fingerprint image identification can be carried out fingerprint recognition to discriminate one's identification.
At present, existing intelligent image disposal route generally is divided into two steps, and first step is to carry out image characteristics extraction, and purpose is that the low-dimensional that obtains image is represented, so that follow-up discriminator is handled; Second step uses various sorters such as arest neighbors (KNN) to the feature after extracting, support vector machine (SVM) is classified, for example Xiaogang Wang and Xiaoou Tang exist " AUnified Framework for Subspace Face Recognition " (IEEETRANSACTIONS ON PATTERN ANALYSIS AND MACHINEINTELLIGENCE, VOL.26, NO.9, SEPTEMBER 2004pp 1222-1228) in the literary composition a kind of recognition of face framework has been proposed, main thought is to utilize principal component analysis (PCA) and linear discriminant analysis to carry out face characteristic to extract, and carries out recognition of face in conjunction with nearest neighbor classifier then.K.E.Gates exists " Fast and Accurate Face Recognition Using Support Vector Machines " a kind of face identification method based on support vector machine has been proposed in (Proceedings of the 2005 IEEE Computer Society Conference onComputer Vision and Pattern Recognition) literary composition.
Said method has been obtained recognition effect preferably, but also exists a lot of not enough.
At first, method based on principal component analysis (PCA) can not guarantee to find the image low-dimensional that helps classifying most to represent, though linear discriminant analysis attempts to find the image low-dimensional that helps classifying most to represent, but be subjected to the limited restriction of registration number of samples in the practical application, linear discriminant analysis also can not guarantee to find optimum low-dimensional projection pattern.
About sorter, nearest neighbor classifier needs bigger storage demand, and operand is big, is not suitable for real-time application.Support vector machine has good classification capacity, but support vector machine own is two class sorters, and when solving the multiclass problem, such as the recognition of face problem, the support vector number is many, and calculated amount is big, also is not suitable for real-time application.
Summary of the invention
The object of the present invention is to provide a kind of disposal system of intelligent image quick and precisely, another purpose provides a kind of intelligent image disposal route, the characteristics of said system and disposal route are that storage demand is little, recognition speed is fast, the accuracy of identification height can be widely used in real-time industrial application and public utility field.
In order to achieve the above object, technical scheme of the present invention is: this is the intelligent image disposal system fast and accurately, comprises image capture module, image conversion module, data transmission module, calculating and memory module, data outputting module.The effect of the image capture module of this system is to obtain pending image, the effect of image conversion module is that the image that will collect is converted into and calculates and form that memory module can be discerned and handle, the effect of calculating and memory module is the feature of storage registered images and the model of intelligent image disposal route, the execution intelligent image is handled, stores processor result, simultaneously, image capture device being sent instruction controls.The effect of data outputting module is that the result that intelligent image is handled is shown.
The technical scheme of the intelligent image disposal route of invention is: its treatment step is: the present invention is divided into two parts, and the first step is to set up the registration sample set, and second step was to carry out new samples identification.The foundation of " registration sample set " is divided into several steps again, at first obtain similar registration sample original image from image input module, detect through computer object then, the location, pre-service and normalization obtain standardized registration sample number word image, again each standardized registration sample number word image is carried out feature extraction, obtain corresponding one-dimensional characteristic vector representation respectively, the proper vector of all registration samples constitutes registration sample characteristics matrix A.According to classification under each proper vector in the matrix A, calculate a Data Analysis Model again, i.e. a projection matrix W.Discern for new samples, at first obtain image to be identified from image input device, detect through computer object, the location, pre-service and normalization obtain the target image to be identified of standard, and then this image is carried out feature extraction obtain characteristics of image to be identified vector y, calculating the discriminant vector z of image to be identified then by data model, analysis and distinguishing vector z obtains recognition result again.
Beneficial effect of the present invention: this is intelligent image disposal system and disposal route thereof quick and precisely, be characterized in that storage demand is little, recognition speed is fast, the accuracy of identification height, can be widely used in real-time industrial application and public utility field, replacing prior art becomes mainstream technology, reduces equipment cost, numerous field boundarys is closed used obvious economic.
Below in conjunction with drawings and Examples the present invention is made comparisons and to explain.
Description of drawings
Fig. 1 is the disposal system of an intelligent image fast and accurately structural principle block diagram of the present invention;
Fig. 2 is the program flow diagram of the disposal route of intelligent image fast and accurately of the present invention.
Embodiment
With reference to Fig. 1, this is the disposal system of an intelligent image fast and accurately structural principle block diagram of the present invention.
As shown in the figure, this is the intelligent image disposal system quick and precisely, comprises image capture module 1, image conversion module 2, data transmission module 3, computing module 4, computing module 5, data transmission module 6, data outputting module 7.
The information transfering relation of this system is: deliver to image conversion module 2 from the information (image) that image capture module 1 collects, the effect of image conversion module is to obtain pending image, and the image that collects is converted into calculates and form that memory module can be discerned and handle, deliver to and calculate and memory module, the effect of memory module is the feature of storage registered images, the result of the model of intelligent image disposal route and program and stores processor, the effect of computing module is to carry out intelligent image to handle simultaneously, image capture module is sent instruction control.The effect of data outputting module is that the result that intelligent image is handled is shown.
Described computing module 4 comprises and carries out CPU and the interlock circuit that intelligent image is handled.
Described memory module 5 comprises storer and is stored in wherein Intelligent treatment program and Intelligent treatment model.
The technical scheme of intelligent image disposal route of the present invention, the detailed division of the effect in each step is as follows:
Obtaining digital picture from image input module is meant: comprise video camera from various image input devices, video recorder, camera or the like, obtain image, computer object detects, the location is meant and utilizes certain algorithm to find the image-region that comprises target to be identified fully, the image in this zone is saved as digitized image.Such as recognition of face,,, then find out the positional information and the size information of everyone face, thereby obtain some images that comprise each human face region fully if there is people's face to exist to a width of cloth input picture.
Computing machine pre-service and normalization are meant: utilizes digital image processing techniques that the digital picture of input is carried out various Flame Image Process, comprises translation, and rotation, convergent-divergent, filtering or the like makes the image that obtains at yardstick, angle, grey level histogram or the like is all consistent.Such as recognition of face, according to the eyes positional information of facial image, by rotation, translation, geometric transformations such as convergent-divergent make that the eyes position of all training facial images is identical, and the size of facial image is also identical, and each is opened facial image and all passes through operations such as histogram equalization.
Image characteristics extraction refers to: the gray scale of input picture or colouring information, or image carried out mathematic(al) manipulation,
The result is expressed as an one-dimensional vector with an input picture, and this vector can be represented the feature of this image.Such as directly with the half-tone information of image as feature, with image according to row or be listed as preferential order image array is expressed as an one-dimensional vector.
Registration sample characteristics matrix A is meant: the proper vector of the corresponding registration sample number word image of each row vector among the A.
Projection matrix W is meant: registration sample characteristics matrix A is carried out svd obtain S
M, U
M, V
MS wherein
MBe the singular value diagonal matrix of the bigger singular value correspondence of svd result's the preceding M of matrix A, U
M, V
MBe respectively the left singular vector matrix and the right singular vector matrix of the bigger singular value correspondence of svd result's the preceding M of matrix A, M is selected according to practical application voluntarily by the user.If diagonal matrix S
MI diagonal element S
M iNon-vanishing, V so
MIn i row V
M iAll elements is all divided by S
M iIf, diagonal matrix S
MI diagonal element S
M iBe zero, V so
MIn i row V
M iConstant, go on foot the matrix V that operation obtains upgrading through this
MAgain the row that belong to the column vector correspondence of i kind attribute in the matrix A number are put into a set and be designated as S
i, U
MBelong to S set middle capable number
iCapable vector add up and obtain a new row vector, this row vector is capable as the i of a new matrix D.The diagonal matrix K that to construct a size again be M * M, make is: i the diagonal element of K calculates according to expression:
α wherein, β are the non-negative constants for example 1.0,1.0 of user preset.Compute matrix V again
M, K, the product of the transposed matrix of matrix D obtain new matrix W, i.e. a projection matrix.
The discriminant vector z that calculates image to be identified by data model is meant: according to following expression formula computational discrimination vector z:
z=W
Ty
Analysis and distinguishing vector z obtains result and is meant: classification c calculates according to following expression formula under the input picture:
With reference to Fig. 2, this is the program flow diagram of the disposal route of intelligent image fast and accurately of the present invention.
As shown in the figure, the program circuit of intelligent image disposal route fast and accurately, its step is as follows: from [starting the intelligent image disposal system] 101, enter [initialization is also controlled image acquisition, converting apparatus] 102, does [model data exist? ] 103 judge, if not, then enter [setting up analytical model] 104, [receiving registered images and attribute thereof] 105, [target detection, location and normalization pre-service] 106, [target's feature-extraction] 107, [setting up feature database] 108, [finding the solution analytical model] 109, [the preservation model] 110 of model computation module; If, then enter [being written into analytical model] 111, [receiving image to be identified] 112, [target detection in the picture recognition module, location and normalization pre-service] 113, [target image feature extraction] 114, [recognition objective attribute] 115, enter then and [show, saving result, the control image capture device] 116, separate two-way, does one the tunnel enter [ends with system operation? ] 117 judge, if then [end] 118; If not, then return [receiving image to be identified] 112; [initialization is also controlled image acquisition, converting apparatus] 102 then returned on another road of [showing saving result, control image capture device] 116.
Below three embodiment to the present invention quick and precisely intelligent image disposal system disposal route be illustrated.
Example one: computer face automatic recognition system
Whole invention implementation procedure is as follows:
1), sets up people's face tranining database.From facial image database, select L people's N to open people's face picture, everyone has N/K to open, then these pictures are carried out pre-service, at first carrying out people's face detects, then each facial image is carried out craft or automatic human eye location, according to the human eye positioning result facial image is carried out yardstick then, angle, normalization, make that the eyes position of every facial image is identical, the facial image size is also identical, a facial image just is expressed as a matrix, element in the matrix is exactly the gray-scale value of corresponding each picture element in the image, matrix representation with each facial image correspondence is the capable vector of one-dimensional characteristic that a size is 1 * d then, this row vector is spliced successively by each row vector of the matrix of a facial image correspondence, and the feature row vector of face images is with regard to composing training eigenmatrix A then, and size is N * d.
2), calculate projection matrix W.At first matrix A is carried out svd and obtain S
M, U
M, V
MS wherein
MBe the singular value diagonal matrix of the bigger singular value correspondence of svd result's the preceding M of matrix A, U
M, V
MBe respectively the left singular vector matrix and the right singular vector matrix of the bigger singular value correspondence of svd result's the preceding M of matrix A, M is selected according to practical application voluntarily by the user.If diagonal matrix S
MI diagonal element S
M iNon-vanishing, V so
MIn i row V
M iAll elements is all divided by S
M iIf, diagonal matrix S
MI diagonal element S
M iBe zero, V so
MIn i row V
M iConstant, go on foot the matrix V that operation obtains upgrading through this
MGive the numbering of unique a 1~L in the registered images storehouse everyone, the row that the row vector that belongs to i people in the matrix A is corresponding number is put into a set and is designated as T
i, U
MBelong to set T middle capable number
iCapable vector add up and obtain a new row vector, this row vector is capable as the i of a new matrix D.The diagonal matrix K that to construct a size again be M * M, make is: i the diagonal element of K calculates according to expression:
α wherein, β gets 1.0,1.0 respectively.Compute matrix V again
M, K, the product of the transposed matrix of matrix D obtain a new matrix W=V
MKD
T, i.e. projection matrix.
3), facial image to be identified carried out computing machine detect and human eye location automatically from moving face, carry out the normalization pre-service then and obtain gray level image, with the pixel features of this image one-dimensional characteristic column vector y that to be expressed as a size be d * 1.
4), calculate dimensionality reduction projection coefficient z=W
Ty
5), carry out recognition of face.Get greatest member among the z, if this element is greater than certain default threshold values g, the call number corresponding class of this element correspondence is as recognition result so:
If this element, thinks then that the prediction facial image does not belong to a certain individual in the training storehouse, is identified as unregistered people's face less than certain default threshold values g.
Example two: computer face expression automatic recognition system
Whole invention implementation procedure is as follows:
1), sets up the human face expression tranining database.Definition list affectionate person class, as laugh at, cry, dejected, sleepy, in surprise, indignation etc. select the N with L kind expression to open facial image from facial image database, and every kind of expression has N/L to open image, then these pictures are carried out pre-service, at first carry out people's face and detect, then each facial image is carried out craft or automatic human eye location, according to the human eye positioning result facial image is carried out yardstick then, angle, normalization makes that the eyes position of every facial image is identical, and the facial image size is also identical, a facial image just is expressed as a matrix, element in the matrix is exactly the gray-scale value of corresponding each picture element in the image, and the matrix representation with each facial image correspondence is the capable vector of one-dimensional characteristic that a size is 1 * d then, and this row vector is spliced successively by each row vector of the matrix of a facial image correspondence, the feature row vector of face images is with regard to composing training eigenmatrix A then, and size is N * d.
2), calculate projection matrix W.At first matrix A is carried out svd and obtain S
M, U
M, V
MS wherein
MBe the singular value diagonal matrix of the bigger singular value correspondence of svd result's the preceding M of matrix A, U
M, V
MBe respectively the left singular vector matrix and the right singular vector matrix of the bigger singular value correspondence of svd result's the preceding M of matrix A, M is selected according to practical application voluntarily by the user.If diagonal matrix S
MI diagonal element S
M iNon-vanishing, V so
MIn i row V
M iAll elements is all divided by S
M iIf, diagonal matrix S
MI diagonal element S
M iBe zero, the i row V among the VM so
M iConstant, go on foot the matrix V that operation obtains upgrading through this
MThe numbering of giving unique a 1~L to every kind of expression in the registered images storehouse, again that the row vector that belongs to i kind expression in the matrix A is corresponding row number are put into a set and are designated as T
i, U
MBelong to set T middle capable number
iCapable vector add up and obtain a new row vector, this row vector is capable as the i of a new matrix D.The diagonal matrix K that to construct a size again be M * M, make is: i the diagonal element of K calculates according to expression:
α wherein, β gets 1.0,1.0 respectively.Compute matrix V again
M, K, the product of the transposed matrix of matrix D obtain a new matrix W=V
MKD
T, i.e. projection matrix.
3), facial image to be identified carried out computing machine detect and human eye location automatically from moving face, carry out the normalization pre-service then and obtain gray level image, with the pixel features of this image one-dimensional characteristic column vector y that to be expressed as a size be d * 1.
4), calculate dimensionality reduction projection coefficient z=W
Ty
5), carry out human face expression identification.The expression of call number correspondence of getting greatest member correspondence among the z is as recognition result.
Example three: computer face sex automatic recognition system
Whole invention implementation procedure is as follows:
1), sets up the face gender tranining database.From facial image database, select N to open facial image with L kind sex, every kind of sex has N/L to open image, then these pictures are carried out pre-service, at first carrying out people's face detects, then each facial image is carried out craft or automatic human eye location, according to the human eye positioning result facial image is carried out yardstick then, angle, normalization, make that the eyes position of every facial image is identical, the facial image size is also identical, a facial image just is expressed as a matrix, element in the matrix is exactly the gray scale or the color value of corresponding each picture element in the image, matrix representation with each facial image correspondence is the one-dimensional characteristic column vector that a size is d * 1 then, and this column vector is spliced successively by each column vector of the matrix of a facial image correspondence, and the characteristic series vector of face images is with regard to composing training eigenmatrix A then, size is N * d, the characteristic series vector of the corresponding facial image of each column vector of A.
2), calculate projection matrix W.At first matrix A is carried out svd and obtain S
M, U
M, V
MS wherein
MBe the singular value diagonal matrix of the bigger singular value correspondence of svd result's the preceding M of matrix A, U
M, V
MBe respectively the left singular vector matrix and the right singular vector matrix of the bigger singular value correspondence of svd result's the preceding M of matrix A, M is selected according to practical application voluntarily by the user.If diagonal matrix S
MI diagonal element S
M iNon-vanishing, V so
MIn i row V
M iAll elements is all divided by S
M iIf, diagonal matrix S
MI diagonal element S
M iBe zero, the i row V among the VM so
M iConstant, go on foot the matrix V that operation obtains upgrading through this
MThe numbering of giving unique a 1~L to every kind of sex in the registered images storehouse, again that the row vector that belongs to i kind sex in the matrix A is corresponding row number are put into a set and are designated as T
i, U
MBelong to set T middle capable number
iCapable vector add up and obtain a new row vector, this row vector is capable as the i of a new matrix D.The diagonal matrix K that to construct a size again be M * M, make is: i the diagonal element of K calculates according to expression:
α wherein, β gets 1.0,1.0 respectively.Compute matrix V again
M, K, the product of the transposed matrix of matrix D obtain a new matrix W=V
MKD
T, i.e. projection matrix.
3), facial image to be identified carried out computing machine detect and human eye location automatically from moving face, carry out the normalization pre-service then and obtain gray level image, with the pixel features of this image one-dimensional characteristic column vector y that to be expressed as a size be d * 1.
4), calculate dimensionality reduction projection coefficient z=W
Ty
5), carry out face gender identification.The sex of call number correspondence of getting greatest member correspondence among the z is as recognition result.
Though the present invention describes with reference to the above embodiments; but those of ordinary skill in the art; will be appreciated that above embodiment is used for illustrating the present invention; should understand and wherein can make variations and modifications and do not break away from the present invention in a broad sense; so be not as limitation of the invention; as long as in connotation scope of the present invention, to the variation of above-described embodiment, the protection domain that distortion all will fall into claim of the present invention.
Claims (8)
1. intelligent image disposal system fast and accurately, comprise a Computerized image processing system, it is characterized in that: this disposal system comprises image capture module (1), image conversion module (2), data transmission module (3), computing module (4), stores processor module (5), data transmission module (6), data outputting module (7);
The information transfering relation of this system is: deliver to image conversion module 2 from the information (image) that image capture module 1 collects, the effect of image conversion module is to obtain pending image, and the image that collects is converted into calculates and form that memory module can be discerned and handle, deliver to and calculate and memory module, the effect of memory module is the feature of storage registered images, the result of the model of intelligent image disposal route and program and stores processor, the effect of computing module is to carry out intelligent image to handle simultaneously, image capture module is sent instruction control.The effect of data outputting module is that the result that intelligent image is handled is shown.
2. according to claim 1 intelligent image disposal system fast and accurately, it is characterized in that described calculating and stores processor module .4 comprise the CPU that carries out intelligent image and handle, are stored in the Intelligent treatment program in the various storeies.
3. be used for the disposal route of intelligent image disposal system fast and accurately, it is characterized in that: the step of this method is: the 1st step: if analytical model is written into, then carried out for the 4th step, otherwise carry out 2 steps of the mat woven of fine bamboo strips; 2 steps of the mat woven of fine bamboo strips: obtain registered images and attribute thereof, and carry out target detection, the location, normalization pre-service and feature extraction obtain a sample characteristics matrix; 3 steps of the mat woven of fine bamboo strips: calculate projection matrix and preserve also loading system as analytical model according to sample characteristics matrix and sample attribute; The 4th step: receive image to be identified, and carry out target detection, location and normalization pre-service and feature extraction; The 5th step: by analytical model computational discrimination vector; The 6th step: with the call number of greatest member correspondence in the discriminant vector attribute label as target image.
4. intelligent image disposal route according to claim 3, it is characterized in that, described 2 steps of the step mat woven of fine bamboo strips: obtain registered images and attribute thereof, and carry out target detection, the location, normalization pre-service and feature extraction obtain a sample characteristics matrix, are to obtain registered images from the digital picture load module, manually demarcate its property value; Its concrete steps are:
(1) in every registered images, treat recognition objective and detect, and the location; Be meant: in registered images, find out the zone that comprises target to be identified fully, and find out its key point and unique point;
(2) target image of registering being carried out the normalization pre-service is meant: utilize various digital image processing techniques that the digital picture of input is carried out various Flame Image Process, comprise translation, rotation, convergent-divergent, filtering or the like makes the image that obtains at yardstick, angle, grey level histogram or the like are all consistent.Such as recognition of face, according to the eyes positional information of facial image, by rotation, translation, geometric transformations such as convergent-divergent make that the eyes position of all training facial images is identical, and the size of facial image is also identical, and each is opened facial image and all passes through operations such as histogram equalization.
(3) feature extraction is meant: the gray scale of an image or colouring information are expressed as an one-dimensional vector, or image is carried out certain mathematic(al) manipulation is an one-dimensional vector with a graphical representation.
(4) manually demarcate the property value that each registers sample.
5. intelligent image disposal route according to claim 3, it is characterized in that, " the sample characteristics matrix " of described step the in 3 steps is meant: is the character representation of each registration sample a row vector, and these row vectors are arranged in order forms the sample characteristics matrix.Calculating projection matrix W is meant: calculate projection matrix W according to following steps:
(1) the sample characteristics matrix is carried out svd
(2) obtain three matrix S according to decomposition result
M, U
M, V
M, S wherein
MBe the singular value diagonal matrix of the bigger singular value correspondence of svd result's the preceding M of matrix A, U
M, V
MBe respectively the left singular vector matrix and the right singular vector matrix of the bigger singular value correspondence of svd result's the preceding M of matrix A, M is selected according to practical application voluntarily by the user.
(3) upgrade matrix V
MIf: diagonal matrix S
MI diagonal element S
M iNon-vanishing, V so
MIn i row V
M iAll elements is all divided by S
M iIf, diagonal matrix S
MI diagonal element S
M iBe zero, V so
MIn i row V
M iConstant.
(4) the new matrix D of structure: make is, at first to the attribute of registered images from 1~K numbering, wherein K is total attribute classification number, the row that the row vector that belongs to i kind attribute in the matrix A is corresponding number are put into a set and are designated as T
i, U
MBelong to set T middle capable number
iCapable vector add up that to obtain a new row vector capable as the i of a new matrix D, i=1 wherein ..., K.
(5) diagonal matrix K that size is M * M of structure, make is: i the diagonal element of K calculates according to expression:
α wherein, β is for example α of user selected parameter, β gets 1.0,1.0 respectively.
(6) calculate projection matrix W=V
MKD
T
6. intelligent image disposal route according to claim 3 is characterized in that, described step the is calculated in 5 steps and judged that vector is meant by following expression formula design factor:
z=W
Ty。
7. intelligent image disposal route according to claim 3, it is characterized in that described step the is in 6 steps " with the call number of greatest member correspondence in the discriminant vector attribute label as target image " is meant that the attribute c that treats recognition image according to following expression formula discerns:
8. intelligent image disposal route according to claim 3, it is characterized in that, the program circuit of the described disposal route of intelligent image fast and accurately, its step is as follows: from [starting the intelligent image disposal system] 101, enter that [initialization is also controlled image acquisition, converting apparatus] 102, does [model data exist? ] 103 judge, if not, then enter [the setting up analytical model] 104 of model computation module, [receiving registered images and attribute thereof] 105, [target detection, location and normalization pre-service] 106, [target's feature-extraction] 107, [setting up feature database] 108, [finding the solution analytical model] 109, [preservation model] 110; If, then enter [being written into analytical model] 111, [receiving image to be identified] 112, [target detection in the picture recognition module, location and normalization pre-service] 113, [target image feature extraction] 114, [recognition objective attribute] 115, enter then and [show, saving result, the control image capture device] 116, separate two-way, does one the tunnel enter [ends with system operation? ] 117 judge, if then [end] 118; If not, then return [receiving image to be identified] 112; [initialization is also controlled image acquisition, converting apparatus] 102 then returned on another road of [showing saving result, control image capture device] 116.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNA2007100388464A CN101276404A (en) | 2007-03-30 | 2007-03-30 | System and method for quickly and exactly processing intelligent image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNA2007100388464A CN101276404A (en) | 2007-03-30 | 2007-03-30 | System and method for quickly and exactly processing intelligent image |
Publications (1)
Publication Number | Publication Date |
---|---|
CN101276404A true CN101276404A (en) | 2008-10-01 |
Family
ID=39995837
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CNA2007100388464A Pending CN101276404A (en) | 2007-03-30 | 2007-03-30 | System and method for quickly and exactly processing intelligent image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101276404A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102314612A (en) * | 2010-07-01 | 2012-01-11 | 北京中星微电子有限公司 | Method and device for identifying smiling face image and image acquisition equipment |
CN102737249A (en) * | 2011-04-14 | 2012-10-17 | 株式会社日立制作所 | Object identification device and object identification method |
CN108494871A (en) * | 2018-04-08 | 2018-09-04 | 上海微穹信息科技有限公司 | Cloud virtual machine and its image-recognizing method, system and virtual machine thin-client |
CN108932468A (en) * | 2018-04-27 | 2018-12-04 | 衡阳师范学院 | One kind being suitable for psychologic face recognition method |
CN109597945A (en) * | 2011-07-20 | 2019-04-09 | 电子湾有限公司 | Real-time location-aware is recommended |
CN112364675A (en) * | 2020-10-28 | 2021-02-12 | 三维码(厦门)网络科技有限公司 | Off-line reading method and device based on three-dimensional code and image recognition |
-
2007
- 2007-03-30 CN CNA2007100388464A patent/CN101276404A/en active Pending
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102314612A (en) * | 2010-07-01 | 2012-01-11 | 北京中星微电子有限公司 | Method and device for identifying smiling face image and image acquisition equipment |
CN102314612B (en) * | 2010-07-01 | 2016-08-17 | 北京中星微电子有限公司 | The recognition methods of a kind of smiling face's image, identification device and image acquisition equipment |
CN102737249A (en) * | 2011-04-14 | 2012-10-17 | 株式会社日立制作所 | Object identification device and object identification method |
CN102737249B (en) * | 2011-04-14 | 2015-07-08 | 日立产业控制解决方案有限公司 | Object identification device and object identification method |
CN109597945A (en) * | 2011-07-20 | 2019-04-09 | 电子湾有限公司 | Real-time location-aware is recommended |
CN109597945B (en) * | 2011-07-20 | 2023-05-02 | 电子湾有限公司 | Method for generating location-aware recommendations |
CN108494871A (en) * | 2018-04-08 | 2018-09-04 | 上海微穹信息科技有限公司 | Cloud virtual machine and its image-recognizing method, system and virtual machine thin-client |
CN108494871B (en) * | 2018-04-08 | 2022-02-22 | 上海微穹信息科技有限公司 | Cloud virtual machine, image identification method and system thereof and virtual machine thin client |
CN108932468A (en) * | 2018-04-27 | 2018-12-04 | 衡阳师范学院 | One kind being suitable for psychologic face recognition method |
CN108932468B (en) * | 2018-04-27 | 2021-10-12 | 衡阳师范学院 | Face recognition method suitable for psychology |
CN112364675A (en) * | 2020-10-28 | 2021-02-12 | 三维码(厦门)网络科技有限公司 | Off-line reading method and device based on three-dimensional code and image recognition |
CN112364675B (en) * | 2020-10-28 | 2022-11-01 | 三维码(厦门)网络科技有限公司 | Off-line reading method and device based on three-dimensional code and image recognition |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Ismael et al. | Face recognition using Viola-Jones depending on Python | |
CN113657425B (en) | Multi-label image classification method based on multi-scale and cross-modal attention mechanism | |
CN105447532A (en) | Identity authentication method and device | |
CN111695392B (en) | Face recognition method and system based on cascade deep convolutional neural network | |
CN107451545B (en) | The face identification method of Non-negative Matrix Factorization is differentiated based on multichannel under soft label | |
Dornaika et al. | Age estimation in facial images through transfer learning | |
CN102938065A (en) | Facial feature extraction method and face recognition method based on large-scale image data | |
Elharrouss et al. | Pose-invariant face recognition with multitask cascade networks | |
CN110598603A (en) | Face recognition model acquisition method, device, equipment and medium | |
CN101276404A (en) | System and method for quickly and exactly processing intelligent image | |
Sumi et al. | Human gender detection from facial images using convolution neural network | |
Benkaddour et al. | Human age and gender classification using convolutional neural network | |
Amaro et al. | Evaluation of machine learning techniques for face detection and recognition | |
Andiani et al. | Face recognition for work attendance using multitask convolutional neural network (MTCNN) and pre-trained facenet | |
CN103942572A (en) | Method and device for extracting facial expression features based on bidirectional compressed data space dimension reduction | |
CN110287973B (en) | Image feature extraction method based on low-rank robust linear discriminant analysis | |
CN112257600A (en) | Face recognition method and system | |
Gupta et al. | HaarCascade and LBPH Algorithms in Face Recognition Analysis | |
Ma et al. | Bottleneck feature extraction-based deep neural network model for facial emotion recognition | |
Arjun et al. | Multimodal biometric recognition: fusion of modified adaptive bilinear interpolation data samples of face and signature using local binary pattern features | |
Wang et al. | Histogram feature-based Fisher linear discriminant for face detection | |
Hachad et al. | Student's attendance management using deep facial recognition | |
CN113947780A (en) | Sika deer face recognition method based on improved convolutional neural network | |
Huang et al. | Age-puzzle facenet for cross-age face recognition | |
Telli et al. | A novel multi-level pyramid co-variance operators for estimation of personality traits and job screening scores |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Open date: 20081001 |