CN114758356A - Method and system for recognizing cow lip prints based on local invariant features - Google Patents
Method and system for recognizing cow lip prints based on local invariant features Download PDFInfo
- Publication number
- CN114758356A CN114758356A CN202210310627.1A CN202210310627A CN114758356A CN 114758356 A CN114758356 A CN 114758356A CN 202210310627 A CN202210310627 A CN 202210310627A CN 114758356 A CN114758356 A CN 114758356A
- Authority
- CN
- China
- Prior art keywords
- cattle
- image
- lip print
- lip
- cow
- 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
- 238000000034 method Methods 0.000 title claims abstract description 60
- 241000283690 Bos taurus Species 0.000 claims abstract description 231
- 238000012795 verification Methods 0.000 claims abstract description 8
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 238000009395 breeding Methods 0.000 claims description 16
- 230000001488 breeding effect Effects 0.000 claims description 16
- 238000004422 calculation algorithm Methods 0.000 claims description 13
- 238000000605 extraction Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 8
- 238000009826 distribution Methods 0.000 claims description 5
- 238000003860 storage Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 230000006870 function Effects 0.000 claims description 3
- 244000144972 livestock Species 0.000 abstract description 18
- 241001465754 Metazoa Species 0.000 description 20
- 239000013598 vector Substances 0.000 description 13
- 238000001514 detection method Methods 0.000 description 10
- 238000012360 testing method Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 6
- 238000011161 development Methods 0.000 description 5
- 230000018109 developmental process Effects 0.000 description 5
- 239000011159 matrix material Substances 0.000 description 5
- 230000004044 response Effects 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 244000309464 bull Species 0.000 description 4
- 238000010276 construction Methods 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 201000010099 disease Diseases 0.000 description 3
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 3
- 235000013305 food Nutrition 0.000 description 3
- 241000894007 species Species 0.000 description 3
- 210000004243 sweat Anatomy 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 239000008267 milk Substances 0.000 description 2
- 210000004080 milk Anatomy 0.000 description 2
- 235000013336 milk Nutrition 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 239000003973 paint Substances 0.000 description 2
- 210000003462 vein Anatomy 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 101710149792 Triosephosphate isomerase, chloroplastic Proteins 0.000 description 1
- 101710195516 Triosephosphate isomerase, glycosomal Proteins 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 239000011324 bead Substances 0.000 description 1
- 235000015278 beef Nutrition 0.000 description 1
- 230000003542 behavioural effect Effects 0.000 description 1
- 230000001149 cognitive effect Effects 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000004040 coloring Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 235000012054 meals Nutrition 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000000059 patterning Methods 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000033772 system development Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a method and a system for recognizing cow lip prints based on local invariant features, wherein the method comprises the following steps: acquiring a cow lip print image in real time, and preprocessing the image; extracting feature descriptors from the obtained cow lip print image, storing the feature descriptors in a cow lip print descriptor database, uploading and storing the information of the cow lip print descriptor database in a cow identity information database, and corresponding the feature descriptors in the database with the cow identity information to create an exclusive id number; and establishing an insurance application and verification system, and verifying the cattle insurance according to the exclusive id number. According to the invention, the camera is used for acquiring the lip print image of the cattle, the remote server is used for matching the lip print image to identify the identity information of the cattle, and the accuracy of cattle identity identification in the livestock insurance industry is improved.
Description
Technical Field
The invention relates to the technical field of bovine lip print recognition algorithms, in particular to a bovine lip print recognition method and system based on local invariant features.
Background
Livestock is an important part of agriculture, especially big livestock, and plays an important role in getting rich by farmers. However, some farmers are blindly cultivated and eat meals by the day, and once the farmers encounter natural disasters and disasters or serious epidemic situations, the farmers become overwhelmed. The large livestock insurance belongs to one of the livestock industry insurance, and is insurance taking the life value of livestock as an insurance target.
The cattle insurance policy is an important component of the national policy agricultural insurance system, if cattle die unnaturally due to diseases or external factors such as natural disasters in the breeding process, insurance companies pay insurance money to herdsmen, but due to the lack of cattle identification technology, problems of cheating insurance, missed claims and the like can occur. Meanwhile, the cattle insurance policy becomes an important means for supporting the development of beef cattle and guaranteeing the breeding benefits of governments at all levels, the accurate identification of individual identities promotes the smooth development of cattle insurance, the problem that the claim cattle and the underwriting cattle cannot be completely matched is solved, and the accuracy of the underwriting claim is improved.
Aiming at the identification of cattle, a computer vision can be used for designing an identification system for livestock insurance with high accuracy, strong robustness and lower cost. The system is used for storing the cattle information into a database, and when the cattle identity identification is needed, the target is matched with the cattle identity information stored in the database to find the identity information of the target cattle.
Many international organizations, such as food safety and the world animal health organization, have formally recognized the important value of animal identification and traceability system development and further actively pushed the development of these systems. Among these, identification and traceability of cattle are key to controlling animal safety policies and food production management. In recent years, identification of cattle has been widely used for various applications such as animal registration, traceability, tracking and control of severe diseases, and the like. It also plays an important role in solving the problems of false insurance claims, loss, exchange and the like of cattle and in the process of livestock registration and traceability.
Conventional individual bovine identification methods are classified into several categories, namely Permanent Identification Method (PIM), semi-permanent identification method (SIM), and Temporary Identification Method (TIM). PIM-based techniques include tattoos, microchips, ear tips or incisions, branding, and the like. However, the PIM method is an intrusion-based recognition method. The SIM method has been applied to the identification of individual cattle using ID paint and ear tags. In addition, electrical signal-based technologies, Radio Frequency Identification (RFID), and body sketch patterning technologies using paint or dye have been used for bovine identification of TIMs. In indonesia, ear tag based technology is well suited to identify different livestock. In addition, in countries such as the united states, australia, europe, canada, and the united kingdom, the ear tag built-in radio frequency identification technology is also widely used in applications for registration, traceability, and identification of livestock. However, the ear tag of cattle is also eventually damaged and may quickly become fraudulent and copied due to prolonged use. In the direction of cattle identification, the cattle can also be identified by using the crushing colors of different varieties. However, the bovine body surface coloring process requires skilled paintability to obtain a better image pattern. Therefore, the recognition effect of the conventional individual cattle recognition method is mostly unsatisfactory.
The conventional animal identification method has limitations in the identification of cattle, such as efficiency, economy, non-invasiveness, low cost, and scalability. It also has serious traceability issues such as identifying lost, exchanged and false insurance claims. In the conventional animal recognition system and the frame-based animal recognition system, there are also serious problems in the prevention and control of disease outbreak, health management, and registration of a large number of animal populations. In addition, all traditional animal identification techniques, manual marking methods such as embedded microchips and RFIDs, can also be duplicated, fraudulent, and not verifiable. Therefore, there is a need to design and develop an automatic, non-invasive, low-cost, robust system for identifying individual cows using computer vision based on animal biometrics. Visual animal biometrics is performed between computer vision, pattern recognition, image processing, and cognitive science. Animal-based visual recognition systems or frameworks are applied to identify and classify different species or individuals on different scales. The recognition system can also detect the occurrence and changes in behavior of different species and calculate morphological patterns, as well as calculate inter-individual and intra-individual variation over time. Due to the diversity in behavioral analysis and biometric testing of animals, it is becoming more widely used. Animal biometric identification techniques have a great impact on the identification of animals or species and are of great interest for the development of new methods based on computer vision. In the prior art, cattle dermatoglyphs (i.e., ridges, beads, and whiskers) are shown on the lip print image, and the lip prints of each cattle are unique. The lip print image recognition of a cow is very similar to the human fingerprint minutiae recognition. Thus, the identification of cow lip print images is an animal biometric identification method suitable for livestock identification, and the identification of individual animals can be successfully used so far, which provides a better solution to the main problems of the former cow identification methods.
Be applied to the livestock insurance field with the individual recognition of livestock based on the lip line, research and development insurance service platform interface solves the difficult trade pain point of herdsman's insurance. In the livestock insurance industry, the identification of livestock is the most critical link and is also the key problem which restricts the access and development of the insurance industry in the grassland animal husbandry at present. The accuracy, uniqueness and feasibility of the identity recognition are important indexes of individual identity recognition in the livestock insurance industry. The traditional animal identification in China has methods such as ear tags, RFID and the like, and the methods have problems, such as influence on cattle or low service life. The lip line of the cattle is similar to the fingerprint of a human body, does not change along with the age of the cattle and the change of the feeding environment, has better identity identification characteristic, utilizes the lip line image of the cattle to identify the identity of the cattle, solves the problem of difficult identification of the cattle in insurance, and has the advantages of high identification accuracy, easy operation, low cost and the like.
Disclosure of Invention
The invention aims to provide a cattle lip print recognition method and system based on local invariant features.
In order to solve the technical problems, the technical scheme of the method and the system for identifying the ox lip print based on the local invariant features is as follows:
in a first aspect, the embodiment of the invention discloses a method for identifying bovine lip prints based on local invariant features, which comprises the following steps:
step 1: acquiring a cow lip print image in real time, and preprocessing the image;
and 2, step: extracting a characteristic descriptor from the obtained cow lip print image, storing the characteristic descriptor in a cow lip print descriptor database, uploading and storing the information of the cow lip print descriptor database into a cow identity information database, and corresponding the characteristic descriptor and the cow identity information in the database and creating an exclusive id number;
and 3, step 3: and establishing an insurance application and verification system, and verifying the cattle insurance according to the exclusive id number.
In any of the above schemes, preferably, the cattle identity information database includes ear tag id, age, breed, sex and pasture information of each cattle, and the cattle lip print descriptor database includes lip print images of each cattle and a plurality of corresponding 64-dimensional feature descriptors.
In any of the above schemes, preferably, when the cattle insurance is verified, the cattle lip print image to be verified and the id number of the cattle are uploaded, the rear end performs feature matching on the cattle lip print image and the feature descriptor stored in the id number, and finally, whether the cattle is the same cattle is evaluated according to a threshold value, and if the cattle is the same cattle, the identity information of the cattle is output.
Preferably in any one of the above schemes, the insuring comprises collecting, uploading and storing of image information and cattle breeding information, a camera of a handset collects images and uploads the images, the cattle breeding information comprises the age of the cattle and cattle farm information affiliated to the cattle, the cattle farm information is input and uploaded to a server through the handset, and the cattle identity information database is stored.
In any of the above schemes, preferably, the application and verification functions are integrated on a webpage, the webpage can be accessed by using an intelligent device, the front-end webpage is connected with the back-end service, the front end sends the image and the related instruction to the back-end server, and the back-end server processes the front-end instruction by using a cow lip print recognition algorithm and returns the result.
In any of the above schemes, preferably, the method for recognizing the ox lip print based on the local invariant features further includes compensating for breeding the cattle, acquiring an image by a camera of the handset after the cattle has an accident, selecting the cattle to be claimed and uploading the cattle identity to the server, inputting the image into the feature extraction model by the server, extracting the features, retrieving the existing features of the claimed cattle from the database, sending the features into the decision maker, verifying whether the uploaded cattle lip print image is consistent with the claimed cattle identity, and sending the result to the handset.
In any of the above aspects, preferably, the pretreatment comprises: the image is processed by CLAHE, the number of pixels in different gray scale ranges is approximately the same, when the gray scale histogram of the image is completely and uniformly distributed, the entropy and the contrast of the image are maximum,
In any of the above aspects, preferably, the pretreatment further comprises: in the actual processing of the digital image, a discrete form needs to be introduced, and the calculation mode of the discretization histogram equalization is as follows:
wherein the variable n is the total number of pixels in the image; the variable k is the number of gray levels, and the number of gray levels of a common RGB image is 0-255.
In any of the above embodiments, it is preferable that the variable r has a size ranging from 0 to 1, 0 representing black and 1 representing white.
Compared with the prior art, the invention has the following beneficial effects:
the camera is used for collecting the lip print image of the cattle, the remote server is used for matching the lip print image to identify the identity information of the cattle, and the accuracy of cattle identity identification in the livestock insurance industry is improved.
In a second aspect, a system for recognizing bovine lip print based on local invariant features comprises:
the acquisition module is used for acquiring the cattle lip print image in real time and preprocessing the image;
the processing module is used for extracting the feature descriptors from the obtained cattle lip print images, storing the feature descriptors into a cattle lip print descriptor database, uploading and storing the information of the cattle lip print descriptor database into a cattle identity information database, and corresponding the feature descriptors in the database with the cattle identity information and creating an exclusive id number;
and the establishing module is used for establishing an application and verification system and verifying the cattle insurance according to the exclusive id number.
Compared with the prior art, the invention has the following beneficial effects: the camera is used for collecting the lip print image of the cattle, the remote server is used for matching the lip print image to identify the identity information of the cattle, and the accuracy of cattle identity identification in the livestock insurance industry is improved.
Drawings
The drawings are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification.
FIG. 1 is a schematic diagram of an application claims system.
Fig. 2 is an image storage flowchart.
Figure 3 CLAHE processing of a bovine lip print image.
Fig. 4 gaussian difference pyramid.
FIG. 5 locates spatial key points.
FIG. 6 Key Point principal direction assignments.
FIG. 7 Key points and principal directions of the bovine lip print image.
FIG. 8 is a schematic diagram of feature descriptor generation.
FIG. 9 is a diagram of an example bovine lip print dataset.
Fig. 10 shows the matching result of the same ID lip print image.
Fig. 11 shows the matching results of different ID lip print images.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
A method for recognizing cow lip print based on local invariant features comprises the following specific contents:
the design of the cattle lip print recognition model, and the cattle lip print recognition method comprises two stages, namely a training stage and a testing stage. In the training phase, a lip print image of the cow is acquired using a camera. And after the resolution of the image is adjusted, performing contrast-limited adaptive histogram equalization (CLAHE) to obtain the feature-enhanced bovine lip print image. Then, using an accelerated robust feature (SURF) feature extraction algorithm to generate 64-dimensional feature descriptors for the clearest image of each cow, and storing each feature descriptor in a cow lip print database; in the testing stage, the steps are repeated, after the feature descriptors of the tested cow lip print image are obtained, the KNN classifier is used for matching the descriptors of the tested cow lip print image with all the descriptors in the cow lip print database, and finally, the recognition result is output.
And establishing an application subsystem, wherein the application subsystem comprises the acquisition, uploading and storage of image information and cattle breeding information. The camera of the handset is used for collecting the ox lip print image and the breeding information. The cattle breeding information comprises the age of the cattle, the cattle farm affiliated to the cattle and other information, is input through the handset, is uploaded to the server, and is stored in the cattle identity information database. As shown in the schematic diagram of the application claims system of fig. 1.
The device is used for uploading image information, the camera of the image acquisition handset acquires images, data acquisition of the device is from a cattle farm and a milk station of a Canhal right flag and a Turmer left flag, and the data acquisition is performed under the conditions of different scenes, angles, illumination and the like. And uploading the ox lip print image to a database to serve as a standard for verification in the claim settlement stage. As shown in the image storage flowchart of fig. 2.
The method comprises the steps of collecting breeding information and recording basic identity information including birth time, age and the like of the cattle. And completing the entry of basic information of the cattle, entering the basic information through the handset, and storing the information into a database information table.
The method comprises the steps that a claim settlement subsystem is obtained, after an accident occurs to a cow, a camera of a handset collects a cow lip print image, then the identity of the cow to be claimed is selected and uploaded to a server, the server inputs the image into a cow lip print recognition model for feature extraction, existing features of the claim settlement cow are searched from a database and sent to a decision maker to verify whether the uploaded cow lip print image is consistent with the claim settlement cow identity or not, and the result is sent to a front-end webpage or the handset.
The method comprises the steps of acquiring image information in real time, collecting a cattle lip print image through a camera of a handset after a cattle accident occurs, and uploading an ID number and a picture address of a cattle to be claimed to a server.
And (3) extracting the features of the bovine lip print, inputting the obtained image into a bovine lip print recognition model by the server for feature extraction, converting the image information into 64-dimensional vectors, retrieving the existing features of the claimed bovine from a database, comparing the similarity of the 64 feature vectors of the bovine lip print by using a nearest neighbor classifier (KNN), and taking the feature vector with the highest similarity to the feature vector of the claimed bovine as a matching result. And inquiring whether the identity information of the feature vector is the cattle to be claimed, and outputting 'yes' or 'no'.
For better understanding of the above technical solutions, the technical solutions of the present invention will be described in detail below with reference to the drawings and the detailed description of the present invention.
Example (b):
the invention relates to a method for recognizing cow lip prints based on local invariant features, which comprises three aspects of designing an identity recognition model, designing a protection subsystem and designing a claim settlement subsystem respectively.
Identity recognition model design
1. Bovine lip print image preprocessing
The shot cow face image contains the cow lip line part, and the lip line part needs to be cut out. Because the shooting distances of the cow head photos are different, the cut cow lip print images have different resolutions, and the accuracy of lip print identification is influenced. The lip print image resolution was assigned to 300 x 300 using three bilinear interpolations.
The image is then processed using CLAHE to make the number of pixels in different gray scale ranges approximately the same. When the image gray level histogram is completely uniformly distributed, the entropy and contrast of the image are maximum. The formula satisfying this condition is shown in formula 1.
In the formula 1, the variable r is the gray level of a pixel in an original image, and after normalization processing is carried out, the value ranges from 0 to 1, 0 represents black, and 1 represents white; the variable S is the pixel gray level in the transformed image; probability density pr(r) is the distribution of the gray levels of the original image. The formula is continuous and the digital imageIn practical processing, a discrete form needs to be introduced, and a discretized histogram is equalized as shown in formula 2.
In formula 2, the variable n is the total number of pixels in the image; the variable k is the number of gray levels, and the number of gray levels of a common RGB image is 0-255. As shown in fig. 3, the CLAHE process of the cattle lip pattern image makes the image gray distribution uniform and the features are easier to identify.
2. Accelerated robust feature (SURF) extraction
2.1 construction of Hessian matrix
The pyramid image of SURF construction is very different from SIFT because these differences speed up its detection. SIFT uses DOG images and SURF uses Hessian matrix determinant approximation images. Firstly, the Hessian matrix of a certain pixel point in the image is shown in formula 3.
Wherein L isxx、Lxy、LyyThe second derivative of the gaussian filtered image g (σ) in each direction.
In order to find out the feature points in the image, the original image needs to be transformed, and the transformation map is formed by the approximate values of the Hessian matrix determinant of each pixel of the original image. See equation 4, where 0.9 is an empirical value for algorithm practice.
det(Happrox)＝LxxLyy-(0.9Lxy)2 (4)
2.2 construction of Gaussian pyramid
In order to make the scale show the continuity of the traditional image pyramid, Gaussian filtering is added on the basis of simple down-sampling. As shown in fig. 4, a gaussian difference pyramid (DoG) is formed by using different scale kernel parameters σ to perform gaussian blurring on one image of each layer of the bovine lip vein image pyramid.
Where σ is the standard deviation of the normal distribution, and the larger the σ value, the more blurred the image. r is the blur radius, which refers to the distance of the template element from the center of the template. If the size of the two-dimensional template is (m × n), the gaussian calculation formula corresponding to the element (x, y) on the template is shown in formula 5:
in the formula, m and n represent the dimension of the Gaussian template and are determined by sigma. (x, y) represents the pixel position of the image. σ is a scale space factor, and smaller values indicate that the image is smoothed less, and the corresponding scale is smaller.
Compared with the Gaussian pyramid construction process of the above algorithm, the SURF algorithm speed is improved. In the SURF algorithm, the image size of each group is not the same, and the next group is a down-sample (1/4 size) of the previous group of images. In the several images within each group, their sizes are the same, except that they use a different scale σ. Also in the process of blurring, their gaussian template size is always constant, only the scale σ changes. For SURF algorithms, the size of the image is always constant, only the size of the gaussian blur template is changed, of course, the scale σ is also changed.
2.3 extracting Key points
Each pixel processed by the hessian matrix is compared with all its neighbors. As shown in fig. 5 for extracting key points, the middle detection point is compared with 26 points, namely 8 adjacent points at the same scale, 9 points corresponding to the upper adjacent scale and 9 points corresponding to the lower adjacent scale, so as to ensure that extreme points are detected in both the scale space and the two-dimensional image space.
And obtaining the scale image of the key point according to the scale sigma of the key point. The calculation of the scaled image is shown in equation 4. In the formula, I (x, y) represents an element on the image I, and G (x, y, σ) represents a corresponding element on the gaussian image pyramid in formula 3.
L(x,y)＝G(x,y,σ)×I(x,y) (6)
2.4 Key Point principal Direction Allocation
In SURF, in order to ensure rotation invariance, a gradient histogram is not counted, but a Harr wavelet feature in the feature point field is counted. Taking a feature point as a center, calculating the sum of Haar wavelet responses of all points in a 60-degree fan in x (horizontal) and y (vertical) directions (the side length of the Haar wavelet is 4s) in a neighborhood with the radius of 6s (s is the scale value of the feature point), giving Gaussian weight coefficients to the response values to make the response contribution close to the feature point large and the response contribution far away from the feature point small, then adding the responses in the 60-degree range to form a new vector, traversing the whole circular region, and selecting the direction of the longest vector as the main direction of the feature point. In this way, the principal direction of each feature point is obtained by calculating the feature points one by one. The process is schematically shown in fig. 6.
In the bull lip print image, the key points are extracted and the principal directions are assigned using this algorithm as shown in the key points and principal directions of the bull lip print image of fig. 7. The extraction of the dominant direction may make lip print image matching have rotational invariance.
2.5 generating feature descriptors
A square box is taken around the feature point, the side length of the box is 20s (s is the scale on which the feature point is detected). The orientation of the box is the main orientation detected at 2.4. The box is then divided into 16 subregions, each subregion counting 25 pixels of Haar wavelet features in both the horizontal and vertical directions, where both the horizontal and vertical directions are relative to the principal direction. The Haar wavelet features are the sum of horizontal direction values, the sum of horizontal direction absolute values, the sum of vertical direction values and the sum of vertical direction absolute values. The generated feature descriptor diagram is shown in fig. 8.
Second, the insurance subsystem
The application subsystem comprises the acquisition, uploading and storage of image information and cattle breeding information. The image is gathered to the handheld quick-witted camera of image acquisition, and handheld quick-witted camera gathers ox lip line image and uploads. The cattle breeding information comprises information which is input through a handset and uploaded to a server, and the information is stored in a cattle identity information database. As shown in the schematic diagram of the insurance application claim settlement system in fig. 1, the work flow of the insurance application subsystem is as follows:
after the image is uploaded to a server, converting the image information into 64-dimensional characteristic vector information by using a trained model, and storing the vector information into a MySql database and a store table;
after the image is uploaded to the server, basic information of the cattle, such as the age of the cattle and basic breeding information of the cattle such as a cattle farm owner, belonging to the cattle, is correspondingly stored in a MySql database and an information table.
1. Uploading of image information
1.1 handset camera acquisition image
In the cattle farm and the milk station of the inner Mongolia Czar right flag and the Turmet left flag, photographs of the heads of cattle were collected using 3000 ten thousand pixel slr digital cameras. The lip print area is cropped from the photograph of the bovine head to form a bovine lip print image dataset. When no feed or sweat is shielded at the position of the ox lip line and the camera focuses on the position, the larger the pixel value of the picture is, the clearer the ox lip line is, and the better the recognition effect is.
The acquired bovine lip print images contain blurred images in various covariate forms. As shown in fig. 9, the first row of the figure is a blurred bull lip image and the second row is a sharp bull lip image. As seen from the first row of FIG. 9, four pictures from left to right are sequentially blurred due to sweat reflection, sweat shielding, food shielding and the movement of the head of the cow, and the four pictures are covariates affecting the images of the lip prints of the cow. From all the acquired bovine lip prints, covariate images and blurred bovine lip prints need to be manually filtered. Finally, 475 clear images for lip print identification were obtained, which contained 51 test cows, with different numbers of 5 to 30 animals per cow. FIG. 9 is an exemplary illustration of a bovine lip print data set.
1.2 bovine lip region detection
The method has the advantages that the accurate acquisition of the image of the cow lip print part from the image can greatly improve the accuracy of identity recognition, the target detection technology based on deep learning develops rapidly, and a beneficial basis is provided for improving the cow lip print part detection. At present, a plurality of detection methods are adopted, and a Yolov5 target detection method is adopted in the device to realize the detection function of the lip print part of the cow.
1.3 image upload
Uploading the cattle lip print image to a server, converting the image information into characteristic vector information through a training model, storing the converted characteristic vector information into a store table, and storing the ID number of the cattle, the characteristic vector information of the cattle and the address of the cattle picture in the table.
2. Collection of breeding information
The cattle identity breeding information is acquired through a keyboard, when images of the lip veins of the cattle are shot, basic information of the cattle is stored in an information table in a database by each shot of cattle, and the archived information comprises information such as the ID number of the cattle, a birth ranch, the membership of a farmer, the age of the cattle and the like.
Third, claim settlement subsystem
1. Image information acquisition
When the cattle has an accident and needs claim settlement, the operation of the insurance subsystem is repeated, the image of the face of the cattle is shot by using the hand-held machine, the lip line part of the cattle is detected by using the YOLOv5 target detection model, and the lip line area image is cut to obtain the lip line part image of the accident cattle. And uploading the accident ox lip print image and the ID number to a server. As shown in the schematic diagram of the application claims system of fig. 1.
2. Bovine lip print feature extraction
And extracting the features of the bovine lip print image through an SURF algorithm to form a plurality of 64-dimensional feature vectors and form feature descriptors.
3. KNN classifier
After the feature descriptors of the accident cow lip print image are obtained, matching the descriptor of the accident cow lip print image with all the descriptors in the cow lip print database by using a KNN feature matching algorithm, outputting an identification result, and sending the result to the mobile phone, wherein the identification result is 'yes' or 'no' of the corresponding relation between the accident cow and the uploaded ID number.
In the bovine lip print feature matching, N64-dimensional feature descriptors are extracted from an image of the bovine lip print to be tested by a SURF feature extraction algorithm. And matching the test picture with a database by using KNN, wherein for each feature descriptor of each ID, the K value is 2, and two matched descriptors are obtained from the test picture by using Euclidean distance matching. One is closest, distance is denoted m, and one is next closest, distance is denoted n. The matching degree is judged to be good or bad by using ratio detection, and when m is less than 0.7n, the closest feature descriptor is the correct match. If the number of correctly matching feature descriptors is greater than 10, the test picture is successfully matched with the ID. And when the matching success ID is multiple, taking the ID with the maximum number of correct matching descriptors as a correct match.
In an actual experiment, the number of feature descriptors successfully matched between lip images of the same cow is large, as shown in fig. 10, which shows the matching result of the lip images with the same ID. The number of feature descriptors successfully matched between lip images of different cows is small, as shown in fig. 11, which shows the matching result of lip images with different IDs. Which cow the test image belongs to can be distinguished by the number of feature descriptors.
Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art will understand that various changes, modifications and substitutions can be made without departing from the spirit and scope of the invention as defined by the appended claims. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A cow lip print identification method based on local invariant features is characterized by comprising the following steps: the method comprises the following steps:
step 1: acquiring a cow lip print image in real time, and preprocessing the image;
step 2: extracting feature descriptors from the obtained cow lip print image, storing the feature descriptors in a cow lip print descriptor database, uploading and storing the information of the cow lip print descriptor database in a cow identity information database, and corresponding the feature descriptors in the database with the cow identity information to create an exclusive id number;
and step 3: and establishing an insurance application and verification system, and verifying the cattle insurance according to the exclusive id number.
2. The method for recognizing the ox lip print based on the local invariant features as claimed in claim 1, wherein: the cattle identity information database comprises ear tag id, age, variety, sex and pasture information of each cattle, and the cattle lip print descriptor database comprises lip print images of each cattle and a plurality of corresponding 64-dimensional feature descriptors.
3. The method for recognizing the ox lip print based on the local invariant feature as claimed in claim 2, wherein the method comprises the following steps: when the cattle insurance is verified, uploading a cattle lip print image to be verified and the id number of the cattle, performing feature matching on the cattle lip print image and a feature descriptor stored by the id number at the rear end, finally evaluating whether the cattle is the same head according to a threshold value, and outputting the identity information of the cattle if the cattle is the same head.
4. The method for recognizing the ox lip print based on the local invariant features as claimed in claim 3, wherein the method comprises the following steps: the application includes collection, upload and the storage of image information and ox information of breeding, and handheld machine camera collection image uploads, and the ox information of breeding includes the ox age of ox, the cattle information of affiliated to, types and uploads the server through handheld machine, deposits the ox identity information database.
5. The method for recognizing the ox lip print based on the local invariant features as claimed in claim 4, wherein the method comprises the following steps: the application and verification functions are integrated on the webpage, access can be carried out by using intelligent equipment, the front-end webpage is connected with the back-end server, the front end sends the images and the related instructions to the back-end server, and the back-end server processes the front-end instructions by using a cow lip print recognition algorithm and returns results.
6. The method for recognizing the bovine lip print based on the local invariant features as claimed in claim 5, wherein the method comprises the following steps: the method comprises the steps of selecting a server, inputting the image into a feature extraction model by the server, extracting features, retrieving the existing features of the claim cattle from a database, sending the existing features into a decision maker, verifying whether the uploaded cattle lip print image is consistent with the claim cattle identity, and sending the result to the handheld after the cattle is in an accident.
7. The method for recognizing the bovine lip print based on the local invariant features as claimed in claim 6, wherein the method comprises the following steps: the pretreatment comprises the following steps: the image is processed by CLAHE, the number of pixels in different gray scale ranges is approximately the same, when the gray scale histogram of the image is completely and uniformly distributed, the entropy and the contrast of the image are maximum,
wherein, the variable r is the gray level of the pixel in the original image, and is normalized; the variable S is the pixel gray level in the transformed image; probability density pr(r) is the distribution of the gray levels of the original image.
8. The method for recognizing the ox lip print based on the local invariant features as claimed in claim 7, wherein: the pre-processing further comprises: in the actual processing of the digital image, a discrete form needs to be introduced, and the calculation mode of the discretization histogram equalization is as follows:
wherein, the variable n is the total number of pixels in the image; the variable k is the number of gray levels, and the number of gray levels of a common RGB image is 0-255.
9. The method for recognizing the ox lip print based on the local invariant features as claimed in claim 8, wherein: the variable r ranges in size from 0 to 1, with 0 representing black and 1 representing white.
10. The utility model provides a ox lip print identification system based on local invariant feature which characterized in that: the method comprises the following steps:
The acquisition module is used for acquiring the cattle lip print image in real time and preprocessing the image;
the processing module is used for extracting the feature descriptors from the obtained cattle lip print images, storing the feature descriptors into a cattle lip print descriptor database, uploading and storing the information of the cattle lip print descriptor database into a cattle identity information database, and corresponding the feature descriptors in the database with the cattle identity information and creating an exclusive id number;
and the establishing module is used for establishing an application and verification system and verifying the cattle insurance according to the exclusive id number.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210310627.1A CN114758356A (en) | 2022-03-28 | 2022-03-28 | Method and system for recognizing cow lip prints based on local invariant features |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210310627.1A CN114758356A (en) | 2022-03-28 | 2022-03-28 | Method and system for recognizing cow lip prints based on local invariant features |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114758356A true CN114758356A (en) | 2022-07-15 |
Family
ID=82326667
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210310627.1A Pending CN114758356A (en) | 2022-03-28 | 2022-03-28 | Method and system for recognizing cow lip prints based on local invariant features |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114758356A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115641458A (en) * | 2022-10-14 | 2023-01-24 | 吉林鑫兰软件科技有限公司 | AI (Artificial intelligence) recognition system for breeding of target to be counted and bank wind control application |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111291683A (en) * | 2020-02-08 | 2020-06-16 | 内蒙古大学 | Dairy cow individual identification system based on deep learning and identification method thereof |
| CN114067351A (en) * | 2021-09-02 | 2022-02-18 | 四川楠水农牧科技有限公司 | Image-based cattle uniqueness identification method, device and management system |
-
2022
- 2022-03-28 CN CN202210310627.1A patent/CN114758356A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111291683A (en) * | 2020-02-08 | 2020-06-16 | 内蒙古大学 | Dairy cow individual identification system based on deep learning and identification method thereof |
| CN114067351A (en) * | 2021-09-02 | 2022-02-18 | 四川楠水农牧科技有限公司 | Image-based cattle uniqueness identification method, device and management system |
Non-Patent Citations (1)
| Title |
|---|
| 王锋 等: "基于局部不变特征的牛只个体唇纹识别方法研究", 《黑龙江畜牧兽医》, 20 January 2022 (2022-01-20), pages 1 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115641458A (en) * | 2022-10-14 | 2023-01-24 | 吉林鑫兰软件科技有限公司 | AI (Artificial intelligence) recognition system for breeding of target to be counted and bank wind control application |
| CN115641458B (en) * | 2022-10-14 | 2023-06-20 | 吉林鑫兰软件科技有限公司 | AI identification system for target cultivation to be counted and bank wind control application method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107292298B (en) | Ox face recognition method based on convolutional neural networks and sorter model | |
| CN106778902B (en) | Dairy cow individual identification method based on deep convolutional neural network | |
| Zin et al. | Image technology based cow identification system using deep learning | |
| CN109858435B (en) | Small panda individual identification method based on face image | |
| Yao et al. | Cow face detection and recognition based on automatic feature extraction algorithm | |
| Cai et al. | Cattle face recognition using local binary pattern descriptor | |
| CN107077626B (en) | Non-invasive multi-modal biometric identification system for animals | |
| CN106897675A (en) | The human face in-vivo detection method that binocular vision depth characteristic is combined with appearance features | |
| CN111753697B (en) | Intelligent pet management system and management method thereof | |
| Almaadeed et al. | Partial shoeprint retrieval using multiple point-of-interest detectors and SIFT descriptors | |
| KR102325259B1 (en) | companion animal life management system and method therefor | |
| Matkowski et al. | Giant panda face recognition using small dataset | |
| CN112101076A (en) | Method and device for identifying pigs | |
| Bello et al. | Deep belief network approach for recognition of cow using cow nose image pattern | |
| Kumar et al. | Animal Biometrics | |
| CN108334870A (en) | The remote monitoring system of AR device data server states | |
| CN108090426A (en) | A kind of group rearing pig personal identification method based on machine vision | |
| CN114758356A (en) | Method and system for recognizing cow lip prints based on local invariant features | |
| Potje et al. | Extracting deformation-aware local features by learning to deform | |
| Wang et al. | Pig face recognition model based on a cascaded network | |
| Ahmad et al. | AI-Driven livestock identification and insurance management system | |
| KR102325250B1 (en) | companion animal identification system and method therefor | |
| CN108921006B (en) | Method for establishing handwritten signature image authenticity identification model and authenticity identification method | |
| Kaur et al. | Cattle identification with muzzle pattern using computer vision technology: a critical review and prospective | |
| CN112883915A (en) | Automatic wheat ear identification method and system based on transfer learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |