Non-contact intelligent cattle face recognition method
Through the contactless intelligent cattle face recognition method, fixed cameras and efficient distributed computing, combined with keyframe extraction and K nearest neighbor classifiers, the problems of low efficiency and large errors of traditional cattle recognition methods are solved, and efficient and accurate cattle recognition and intelligent management are achieved.
Patent Information
- Application Number
- CN202411632062.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-15
- Publication Date
- 2025-03-07
AI Technical Summary
Traditional cattle identification methods are inefficient and have high errors, are greatly affected by environmental and behavioral factors, and lack intelligent and real-time data analysis capabilities, resulting in inefficient health monitoring and management.
The contactless intelligent cow face recognition method is adopted, and the video stream is analyzed in real time through fixed cameras and efficient distributed computing. Combined with keyframe extraction, foreground detection and background analysis, the frontal images of cows are screened out, and the K nearest neighbor classifier is used for identity identification.
It realizes efficient and accurate cattle recognition, reduces manual management costs, improves breeding efficiency and cattle health monitoring and management effects, and supports real-time data analysis and intelligent management.
Smart Images
Figure CN119580324A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of animal identification, and in particular relates to a non-contact intelligent cow face identification method. Background Art
[0002] Non-contact intelligent cattle face recognition methods have gradually gained attention in modern animal husbandry, and are mainly used in the fields of individual cattle identification, health monitoring and behavior analysis. With the development of artificial intelligence and computer vision technology, traditional cattle identification methods have gradually failed to meet the needs of modern farming. At present, cattle identification usually relies on traditional methods such as manual marking and ear tags. These methods are not only inefficient, but also easily affected by factors such as the external environment and cattle behavior, resulting in reduced recognition accuracy. The traditional ear tag recognition method has certain limitations. The ear tag may be unrecognizable due to wear, loss or damage, and in some cases, the ears of cattle may be blocked by other objects, thus affecting the accuracy of recognition.
[0003] Manual marking often requires a lot of human resources, especially in large-scale farms. The high error rate of manual identification not only increases management costs, but may also lead to delays in cattle health monitoring and management. Secondly, existing technologies are also insufficient in data processing and analysis capabilities. Many traditional recognition systems use simple image processing algorithms and cannot cope with complex environmental changes, such as lighting, angle changes, and cattle movement, which affects the stability and reliability of recognition. These systems often lack sufficient intelligence and cannot perform real-time data analysis and feedback, resulting in limited applications in cattle health monitoring and behavior analysis. Furthermore, existing technologies are also not perfect in terms of information integration and sharing. Many recognition systems cannot be effectively connected to the breeding management system, resulting in the emergence of data islands, affecting overall management efficiency.
[0004] Traditional recognition methods usually lack in-depth analysis of individual cattle behaviors and cannot identify abnormal cattle behaviors in a timely manner, which affects the prediction and intervention of cattle health status. Therefore, with the continuous expansion of breeding scale and the increasing management needs, there is an urgent need for an efficient, accurate and intelligent cattle face recognition method to improve the management level of modern animal husbandry. The emergence of non-contact intelligent cattle face recognition methods aims to solve the problems brought about by the above traditional technologies. Through advanced computer vision and deep learning algorithms, efficient cattle recognition, real-time data collection and intelligent analysis can be achieved, thereby promoting the intelligent development of animal husbandry. Summary of the invention
[0005] The present invention proposes a non-contact intelligent cow face recognition method, which solves the problems of low efficiency and high error in traditional cow recognition. By fixing the camera above the side of the cowshed trough and using efficient distributed computing to perform real-time analysis and processing of the video stream, stable recognition of cows can be achieved. In addition, the method effectively screens out the front image of the cow through key frame extraction, foreground detection and background analysis, avoiding recognition errors caused by factors such as lighting changes and motion blur, thereby improving the accuracy and real-time performance of recognition.
[0006] The technical solution of the present invention is implemented as follows: a non-contact intelligent cow face recognition method, the method comprising the following steps:
[0007] S1: Place the camera at a fixed position above the cowshed trough side, adjust the viewing angle to cover the cowshed trough area; transmit the video stream output by the camera to the computing device, and the computing device performs real-time video analysis and processing in a highly efficient distributed manner;
[0008] S2: Process the input video stream frame by frame, use the key frame extraction algorithm, combine front and back detection and key frame screening to obtain a stable cow front image sequence; use a high-efficiency distributed processing unit to perform real-time video analysis and processing, set the preset thresholds of the head and back of the detection target, and realize front and back detection; use foreground detection to judge the video shot by the camera and filter out the key frames in the video; set the key frames with adjustable thresholds and controllable range to perform high-precision low-latency detection;
[0009] S3: For the key frame, traverse the image row by row and column by column; calculate the number of pixels with a gray value of 255 in each row and column; when the minimum number of pixels in a row or column is greater than the specified threshold, it is regarded as the background or foreground; when the background or foreground detected in two or more consecutive frames is the same, the current frame is regarded as the background or foreground and recorded; use the background or foreground as a screening condition to extract and process the image, remove the background or foreground, and thus obtain a stable cow front image sequence;
[0010] S4: For the cow frontal image sequence, the cow face posture is analyzed and screened by combining real-time posture detection, edge extraction and connected domain analysis algorithms to obtain the qualified frontal cow face and segment and extract it from the background; for the frontal cow face, feature extraction is used through K nearest neighbor classifier to realize cow face identity recognition.
[0011] Compared with the existing technology, this method has the following significant differences: most traditional cattle identification methods rely on manual identification or simple image processing technology, which is inefficient and easily affected by environmental factors. However, this method can achieve continuous and real-time monitoring and identification by fixing the camera and combining efficient distributed computing, which significantly improves work efficiency. In terms of data processing, the existing technology often adopts a single-threaded or inefficient processing method, resulting in large delays and unable to meet the needs of real-time analysis. This method uses key frame extraction and foreground detection, and by setting an adjustable threshold, it achieves high-precision and low-latency detection, and can quickly respond to changes in cattle behavior. In image processing technology, traditional methods often lack effective processing methods for complex backgrounds and lighting changes, resulting in reduced recognition accuracy. However, this method effectively separates the front image of the cattle through dynamic analysis of the background and foreground, overcoming the problem of background interference. The K nearest neighbor classifier is used for feature extraction and cattle face identity recognition, which further improves the accuracy and robustness of recognition, and has stronger adaptability than the simple recognition algorithm in the existing technology.
[0012] As a preferred implementation, after acquiring a sequence of cow front images, the front cow face is extracted. The extracted cow front image sequence includes a background and a foreground. The cow posture detection is judged by extracting the image, and the sequence is traversed row by row. In each row, the number of foreground pixels screened out and a whole row of pixels greater than a preset threshold are regarded as cow head features, and the rest are regarded as cow backs. Edge extraction is performed: for the cow head area ROI, the four edge pixels of the image are calculated; connected domain analysis is performed: it is determined whether two pixels are in the same connected domain. If the pixel spacing of the connected domain is greater than the preset threshold, it is regarded as a cow head and segmented and extracted.
[0013] As a preferred implementation, the frontal cow face is compared with the cow face library image to obtain the Euclidean distance; the image that best matches the frontal cow face features is taken as the reference cow face, and the forehead width, left and right side areas, and the ratio of the left and right side areas of the reference cow face are calculated; if the forehead width, left and right side areas, and the ratio of the left and right side areas of the frontal cow face fall within the ratio of the forehead width, left and right side areas, and left and right area of the reference cow face, then it is determined to be the same cow face; if the Euclidean distance between the frontal cow face and the reference cow face image is less than a set threshold, then the cow face is the same cow face; or the closest image with a distance within 10 is taken as the reference cow face, and the Euclidean distance between the two cow face images is less than a set threshold, then the cow face is the same cow face.
[0014] As a preferred implementation, after completing the cow face identification, the data is imported into the cloud platform, a real-time query function is set in the cloud platform, and the cloud platform is used to log in to the cloud platform for querying the cattle herd files. The cloud platform query page includes cow face identification, cattle herd files and cattle farm information. The cloud platform query page queries the cattle farm information according to the cattle farm ID, and the cattle farm information includes the name of the cattle farm, the farm photo and the number of cattle in the farm; the cloud platform query page queries the cattle information according to the ear tag number, and the cattle information includes the cattle photo, the name of the cattle shed, the ear tag number, the coat color, the date of birth, the date of admission, the source, the strain, the breed, the sex, the birth weight, the age in months, the current parity, the father number and the mother number information,
[0015] As a preferred implementation, the mobile terminal performs real-time cow face scanning and recognition through the scanning page in the gimbal, extracts the front cow face after successful scanning, compares the front cow face with the cow face library image, and obtains the Euclidean distance; takes the image that best matches the front cow face feature as the reference cow face, calculates the forehead width, left and right side areas, and the ratio of the left and right side areas of the reference cow face; determines that the forehead width, left and right side areas, and the ratio of the left and right side areas of the front cow face fall within the ratio range of the forehead width, left and right side areas, and left and right area of the reference cow face, and determines that it is the same cow face; if the Euclidean distance between the front cow face and the reference cow face image is less than a set threshold, it is the same cow face.
[0016] As a preferred implementation, when the mobile terminal logs into the cloud platform to query the cattle files, the cloud platform will feedback the data of cattle photos, cattle numbers, father numbers, mother numbers, cattle farms, cattle sheds and cattle shed trough locations to the mobile terminal.
[0017] After adopting the above technical scheme, the beneficial effects of the present invention are: through real-time monitoring and efficient identification, the cost of manual management can be greatly reduced, and the breeding efficiency can be improved. Especially in large-scale farms, it can realize the intelligent management of cattle and reduce the investment of human resources. Secondly, the high-precision recognition ability of this method can timely discover the health status of cattle, facilitate early diagnosis and intervention, thereby improving the growth rate and survival rate of cattle and reducing breeding risks. In addition, the ability of real-time data analysis enables breeding managers to obtain the behavioral data of cattle, and then optimize the feeding management strategy and improve the overall breeding efficiency. Finally, with the continuous advancement of technology, this method can be combined with other intelligent systems in the future to form a multifunctional intelligent breeding platform including health monitoring, behavioral analysis, environmental monitoring, etc., to promote the modernization and intelligent development of animal husbandry. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0019] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION
[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0021] Example:
[0022] like Figure 1 As shown, a non-contact intelligent cow face recognition method comprises the following steps:
[0023] S1: Place the camera at a fixed position above the cowshed trough side, adjust the viewing angle to cover the cowshed trough area; transmit the video stream output by the camera to the computing device, and the computing device performs real-time video analysis and processing in a highly efficient distributed manner;
[0024] S2: Process the input video stream frame by frame, use the key frame extraction algorithm, combine front and back detection and key frame screening to obtain a stable cow front image sequence; use a high-efficiency distributed processing unit to perform real-time video analysis and processing, set the preset thresholds of the head and back of the detection target, and realize front and back detection; use foreground detection to judge the video shot by the camera and filter out the key frames in the video; set the key frames with adjustable thresholds and controllable range to perform high-precision low-latency detection;
[0025] S3: For the key frame, traverse the image row by row and column by column; calculate the number of pixels with a gray value of 255 in each row and column; when the minimum number of pixels in a row or column is greater than the specified threshold, it is regarded as the background or foreground; when the background or foreground detected in two or more consecutive frames is the same, the current frame is regarded as the background or foreground and recorded; use the background or foreground as a screening condition to extract and process the image, remove the background or foreground, and thus obtain a stable cow front image sequence;
[0026] S4: For the cow frontal image sequence, the cow face posture is analyzed and screened by combining real-time posture detection, edge extraction and connected domain analysis algorithms to obtain the qualified frontal cow face and segment and extract it from the background; for the frontal cow face, feature extraction is used through K nearest neighbor classifier to realize cow face identity recognition.
[0027] The working principle and workflow of the non-contact intelligent cow face recognition method can be divided into several main steps, including camera installation and video stream acquisition, video frame processing and key frame extraction, image background and foreground separation, and cow face posture analysis and identity recognition. The following is a detailed description of each step.
[0028] S1: Camera installation and video stream acquisition: In this stage, the camera must first be installed above the trough side of the cowshed to ensure that it can cover the entire trough area of the cowshed. During installation, the camera's viewing angle must be adjusted according to the layout of the cowshed and the range of activity of the cattle to obtain the best shooting effect. The camera will collect video streams in real time and transmit these video data to the computing device. The computing device uses an efficient distributed processing method to quickly analyze and process the video stream in real time. The core of this stage is to ensure the stability and clarity of the video stream for subsequent image processing and analysis.
[0029] S2: Video frame processing and key frame extraction: After the video stream is transmitted to the computing device, the system will process the input video stream frame by frame. Using the key frame extraction algorithm, the system is able to extract a sequence of front images of the cow by combining front and back detection and key frame screening. First, the preset thresholds for the head and back of the detection target are set to achieve effective front and back detection. The foreground detection algorithm will judge the captured video and filter out the key frames in the video to ensure that these key frames can represent the front state of the cow. In this process, the key frame setting is adjustable, allowing the user to adjust the threshold and range as needed to achieve high-precision low-latency detection. The purpose of this stage is to ensure that the extracted front image of the cow has sufficient clarity and stability, laying the foundation for subsequent image processing.
[0030] S3: Image background and foreground separation: For the extracted key frames, the system will perform row and column traversal on each frame of the image, and calculate the number of pixels with a grayscale value of 255 in each row and column. By setting a specified threshold, when the minimum number of pixels in a row or column is greater than the threshold, the system regards it as the background or foreground. Next, the system will record the situation where the background or foreground detected in two or more consecutive frames is the same, and regard the current frame as the background or foreground. The core of this process is to screen out stable background and foreground information through dynamic analysis of the image. Through these screening conditions, the system can extract and process the image, remove unnecessary background or foreground, and obtain a stable sequence of cow front images.
[0031] S4: Cow face posture analysis and identity recognition: After obtaining a stable sequence of cow frontal images, the system will combine real-time posture detection, edge extraction and connected domain analysis algorithms to conduct in-depth analysis and screening of the cow face posture. Through these algorithms, the system can identify the frontal cow face that meets the conditions and segment and extract it from the background. Once the cow face image is extracted, the system will further perform feature extraction and use the K nearest neighbor classifier to identify the cow face. The feature extraction process includes analyzing the unique features of the cow face, such as color, texture and shape, which will be used as the basis for classification to confirm the identity.
[0032] In summary, this non-contact intelligent cow face recognition method realizes the complete process from video acquisition to identity recognition through reasonable step design. First, the real-time video stream of the cow is obtained through the camera, and then the video is processed frame by frame to extract the key frames. Then the background and foreground of the image are separated, and finally the identity of the cow face is recognized through posture analysis and feature extraction. Each step of this method has been carefully designed to ensure the accuracy and real-time performance of cow face recognition, providing effective technical support for the management and monitoring of modern animal husbandry.
[0033] After obtaining the frontal image sequence of the cow, the frontal cow face is extracted. The extracted frontal image sequence of the cow includes the background and foreground. The cow posture detection is judged by extracting the image, and the sequence is traversed row by row. In each row, the number of foreground pixels screened out and the whole row of pixels greater than the preset threshold are regarded as the cow head features, and the rest are regarded as the cow's back; edge extraction is performed: for the cow head area ROI, the four edge pixels of the image are calculated; connected domain analysis is performed: determine whether two pixels are in the same connected domain. If the pixel spacing of the connected domain is greater than the preset threshold, it is regarded as a cow's head and segmented and extracted.
[0034] This technical solution determines the posture of the cow by traversing the image line by line and screening the number of foreground pixels. The innovation of this method is that it clearly identifies the features of the cow's head based on line traversal and the number of pixels, while the existing technology often relies on complex machine learning algorithms or manual annotation, resulting in slow processing speed and dependence on a large amount of labeled data. By setting a preset threshold, the system can quickly distinguish the features of the cow's head and the back of the cow. This efficient processing method makes the extraction of the cow's face faster and more accurate. In addition, the edge extraction and connected domain analysis in the implementation method are combined to make the judgment of the cow's head area more accurate, which can effectively reduce the misrecognition rate. In the extraction of image features, the existing technology may use the traditional edge detection method, which lacks comprehensive consideration of the connectivity analysis in the image, and easily leads to blurred edges or feature loss.
[0035] The frontal cow face is compared with the cow face library image to obtain the Euclidean distance; the image that best matches the frontal cow face features is taken as the reference cow face, and the width of the forehead, the area of the left and right sides, and the ratio of the area of the left and right sides of the reference cow face are calculated; if the width of the forehead, the area of the left and right sides, and the ratio of the area of the left and right sides of the frontal cow face fall within the range of the ratio of the width of the forehead, the area of the left and right sides, and the ratio of the area of the left and right sides of the reference cow face, then it is determined to be the same cow face; if the Euclidean distance between the frontal cow face and the reference cow face image is less than a set threshold, then the cow face is the same cow face; or the closest image with a distance within 10 is taken as the reference cow face, and the Euclidean distance between the two cow face images is less than the set threshold, then the cow face is the same cow face.
[0036] This technical solution determines the similarity of cow faces by obtaining the Euclidean distance. The advantage of this method is that it is simple and intuitive to calculate, and can quickly obtain results. Compared with the complex deep learning model commonly used in the prior art for image matching, this method not only reduces the complexity of calculation, but also reduces the dependence on high-performance computing resources. In addition, the implementation method also introduces the calculation of the ratio of the width of the forehead and the area on both sides, which further improves the accuracy and reliability of cow face recognition. The existing technology usually only relies on a single feature point or the overall image similarity for judgment, lacks in-depth analysis of detailed features, and is prone to misjudgment.
[0037] After completing the cow face identification, the data is imported into the cloud platform, and a real-time query function is set in the cloud platform. The cloud platform is logged in to the cloud platform using a mobile terminal to query the cattle herd files. The cloud platform query page includes cow face identification, cattle herd files and cattle farm information. The cloud platform query page queries the cattle farm information according to the cattle farm ID, and the cattle farm information includes the name of the cattle farm, the farm photo and the number of cattle in the farm; the cloud platform query page queries the cattle information according to the ear tag number, and the cattle information includes the cattle photo, the name of the cattle shed, the ear tag number, the coat color, the date of birth, the date of admission, the source, the strain, the breed, the sex, the birth weight, the age in months, the current parity, the father number and the mother number information.
[0038] After the cow face identification is completed, the implementation method of importing the data into the cloud platform reflects the innovation of the current technical solution in data management and information query. The introduction of the cloud platform makes the storage and access of data more convenient. Users can query the herd files in real time through the mobile terminal to obtain the required information. The advantage of this function lies in its efficiency and real-time nature. Users can obtain detailed information of cattle anytime and anywhere, such as ear tag number, coat color, date of birth, etc., which greatly improves management efficiency. However, the existing technology often lacks a centralized data management system, and data storage is scattered. When users query information, they need to search manually, which is time-consuming and error-prone. In addition, the use of the cloud platform can also realize the collaborative management of multiple users, which improves the convenience of information sharing.
[0039] The mobile terminal performs real-time cow face scanning and recognition through the scanning page in the gimbal. After the scan is successful, the front cow face is extracted, and the front cow face is compared with the cow face library image to obtain the Euclidean distance; the image that best matches the front cow face features is taken as the reference cow face, and the forehead width, left and right side areas, and the ratio of the left and right side areas of the reference cow face are calculated; it is determined that the forehead width, left and right side areas, and the ratio of the left and right side areas of the front cow face fall within the ratio range of the forehead width, left and right side areas, and left and right area of the reference cow face, and it is determined to be the same cow face; if the Euclidean distance between the front cow face and the reference cow face image is less than the set threshold, it is the same cow face.
[0040] The function of real-time cow face scanning and recognition on the mobile terminal in the implementation method further enhances the practicality of the system. Users can quickly extract the front cow face information and compare it with the image in the library by scanning the cow face through a mobile device. This convenient operation method can significantly improve work efficiency in practical applications. The existing technology in this regard usually requires more complex equipment configuration and operation procedures, and it is difficult to achieve flexible on-site operation. However, this solution realizes the instant recognition and comparison of cow faces through the application of the mobile terminal, and has a good user experience.
[0041] When the mobile terminal logs in to the cloud platform to query the cattle files, the cloud platform will feedback the data of cattle photos, cattle numbers, sire numbers, dam numbers, cattle farms, cattle sheds and cattle shed trough locations to the mobile terminal. The feedback mechanism of the cloud platform query function can feedback cattle photos, cattle numbers, sire numbers, dam numbers and other information to the mobile terminal. The design of this function enables users to obtain comprehensive information when querying, which greatly facilitates the management and tracking of cattle. The existing technology is often relatively simple in information feedback and lacks the integration of multi-dimensional information, which results in users being unable to obtain a comprehensive view when querying, affecting decision-making efficiency.
[0042] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A non-contact intelligent cow face recognition method, characterized in that: The method comprises the following steps: S1: Place the camera at a fixed position above the cowshed trough side, adjust the viewing angle to cover the cowshed trough area; transmit the video stream output by the camera to the computing device, and the computing device performs real-time video analysis and processing in a highly efficient distributed manner; S2: Process the input video stream frame by frame, use the key frame extraction algorithm, combine front and back detection and key frame screening to obtain a stable cow front image sequence; use a high-efficiency distributed processing unit to perform real-time video analysis and processing, set the preset thresholds of the head and back of the detection target, and realize front and back detection; use foreground detection to judge the video shot by the camera and screen out the key frames in the video; Set keyframes with adjustable thresholds and controllable ranges to perform high-precision, low-latency detection; S3: For the key frame, traverse the image row by row and column by column; calculate the number of pixels with a gray value of 255 in each row and column; when the minimum number of pixels in a row or column is greater than the specified threshold, it is regarded as the background or foreground; when the background or foreground detected in two or more consecutive frames is the same, the current frame is regarded as the background or foreground and recorded; use the background or foreground as a screening condition to extract and process the image, remove the background or foreground, and thus obtain a stable cow front image sequence; S4: For the cow frontal image sequence, the cow face posture is analyzed and screened by combining real-time posture detection, edge extraction and connected domain analysis algorithms to obtain the qualified frontal cow face and segment and extract it from the background; for the frontal cow face, feature extraction is used through K nearest neighbor classifier to realize cow face identity recognition.
2. A non-contact intelligent cow face recognition method as claimed in claim 1, characterized in that: After obtaining the frontal image sequence of the cow, the frontal cow face is extracted. The extracted frontal image sequence of the cow includes the background and foreground. The cow posture detection is judged by extracting the image, and the sequence is traversed row by row. In each row, the number of foreground pixels screened out and the whole row of pixels greater than the preset threshold are regarded as the cow head features, and the rest are regarded as the cow's back; edge extraction is performed: for the cow head area ROI, the four edge pixels of the image are calculated; connected domain analysis is performed: determine whether two pixels are in the same connected domain. If the pixel spacing of the connected domain is greater than the preset threshold, it is regarded as a cow's head and segmented and extracted.
3. A non-contact intelligent cow face recognition method as claimed in claim 1, characterized in that: The frontal cow face is compared with the cow face library image to obtain the Euclidean distance; the image that best matches the frontal cow face features is taken as the reference cow face, and the width of the forehead, the area of the left and right sides, and the ratio of the area of the left and right sides of the reference cow face are calculated; if the width of the forehead, the area of the left and right sides, and the ratio of the area of the left and right sides of the frontal cow face fall within the range of the ratio of the width of the forehead, the area of the left and right sides, and the ratio of the area of the left and right sides of the reference cow face, then it is determined to be the same cow face; if the Euclidean distance between the frontal cow face and the reference cow face image is less than a set threshold, then the cow face is the same cow face; or the closest image with a distance within 10 is taken as the reference cow face, and the Euclidean distance between the two cow face images is less than the set threshold, then the cow face is the same cow face.
4. A non-contact intelligent cow face recognition method as claimed in claim 1, characterized in that: After completing the cow face identification, the data is imported into the cloud platform, and a real-time query function is set in the cloud platform. The cloud platform is logged in to the cloud platform using a mobile terminal to query the cattle herd files. The cloud platform query page includes cow face identification, cattle herd files and cattle farm information. The cloud platform query page queries the cattle farm information based on the cattle farm ID, and the cattle farm information includes the cattle farm name, farm photos and the number of cattle in the farm; the cloud platform query page queries the cattle information based on the ear tag number, and the cattle information includes cattle photos, cattle house name, ear tag number, coat color, date of birth, date of entry, source, strain, breed, gender, birth weight, age in months, current parity, sire number and mother number information.
5. A non-contact intelligent cow face recognition method as claimed in claim 4, characterized in that: The mobile terminal performs real-time cow face scanning and recognition through the scanning page in the gimbal. After the scan is successful, the front cow face is extracted, and the front cow face is compared with the cow face library image to obtain the Euclidean distance; the image that best matches the front cow face features is taken as the reference cow face, and the forehead width, left and right side areas, and the ratio of the left and right side areas of the reference cow face are calculated; it is determined that the forehead width, left and right side areas, and the ratio of the left and right side areas of the front cow face fall within the ratio range of the forehead width, left and right side areas, and left and right area of the reference cow face, and it is determined to be the same cow face; if the Euclidean distance between the front cow face and the reference cow face image is less than the set threshold, it is the same cow face.
6. A non-contact intelligent cow face recognition method as claimed in claim 4, characterized in that: When the mobile terminal logs in to the cloud platform to query the cattle files, the cloud platform feeds back the data of cattle photos, cattle numbers, father numbers, mother numbers, cattle farms, cattle sheds and cattle shed trough locations to the mobile terminal.