Cashmere goat identifying and tracking method and system based on large model

Abstract

The invention provides a large-model-based down producing goat recognition and tracking method and system, and the method comprises the steps: collecting goat face data, and carrying out the fine adjustment training of a large model through the goat face data, and obtaining a first goat face recognition module; monitoring video data of target cashmere goats in a breeding area are input into a first goat face recognition module, and the first goat face recognition module outputs a first goat face recognition result and a second goat face recognition result; wherein the first sheep face recognition result is obtained based on analysis of the static sheep face data, and the second sheep face recognition result is obtained based on analysis of the dynamic sheep face data; fusing the first sheep face recognition result and the second sheep face recognition result to obtain a sheep face recognition result; and tracking the target cashmere goat according to the goat face recognition result. According to the invention, identity recognition is carried out according to static information and dynamic information of the face of the down producing goat, and the accuracy of identity recognition and tracking of the down producing goat can be obviously improved.

Description

Technical Field

[0001] The present invention relates to the field of smart breeding technology, and in particular to a cashmere goat identification and tracking method and system based on a large model. Background Art

[0002] As animal husbandry rapidly develops towards modernization, digitization and intelligence, building farms that integrate big data has become an industry trend. By collecting individual information of livestock, precision breeding is carried out to improve breeding efficiency, reduce costs and achieve healthy breeding. In this context, accurate identification of each livestock has become a prerequisite for collecting individual information.

[0003] Currently, the main method for farms to identify livestock is to wear identification ear tags. However, this method has the problems of low identification efficiency, high labor intensity and frequent maintenance. In addition, ear tags may cause discomfort and pain to livestock, and even cause complications of ear integrity. Therefore, it is urgent to find an efficient and non-invasive livestock identification method.

[0004] In recent years, with the development of deep learning technology, scholars have begun to use these technologies to learn the biological characteristics of livestock in order to accurately identify the livestock. Among various livestock biometric identification methods, facial recognition is considered to be one of the most promising livestock identification methods because it does not require contact with the livestock, achieves efficient identification, and effectively avoids livestock stress behavior.

[0005] However, the current sheep face recognition technology is mainly based on the static features of the sheep face (such as facial features, fur color, facial contour, etc.), but in the actual recognition process, the sheep face is often in an active state, and sheep face recognition based solely on static features is prone to recognition failure or recognition errors. Therefore, it is necessary to further improve the sheep face recognition method to improve the accuracy of sheep face recognition. Summary of the invention

[0006] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a cashmere goat identification and tracking method, system, electronic equipment, computer storage medium and computer program product based on a large model.

[0007] The present invention provides a cashmere goat identification and tracking method based on a large model, the method comprising the following steps: Collect sheep face data, and use the sheep face data to fine-tune the large model to obtain a first sheep face recognition module; wherein the sheep face data includes static sheep face data and dynamic sheep face data; Inputting the monitoring video data of the target cashmere goats in the breeding area into the first goat face recognition module, the first goat face recognition module outputs a first goat face recognition result and a second goat face recognition result; wherein the first goat face recognition result is obtained based on the analysis of the static goat face data, and the second goat face recognition result is obtained based on the analysis of the dynamic goat face data; Obtaining a sheep face recognition result based on the fusion of the first sheep face recognition result and the second sheep face recognition result; The target cashmere goat is tracked according to the sheep face recognition result.

[0008] Optionally, the collecting of sheep face data, and using the sheep face data to fine-tune the large model to obtain a first sheep face recognition module, comprises: When the face of each cashmere goat in the breeding area is in a static state, the goat face images under multiple angles and multiple lighting conditions are photographed, and the static goat face data is extracted from the goat face images; When each cashmere goat in the breeding area is in a facial activity state, a facial action video is shot, and the dynamic goat face data is extracted from the facial action video; The static sheep face data is used to perform preliminary fine-tuning on the large model to learn the basic features of the sheep face; and the dynamic sheep face data is used to further fine-tune the large model after preliminary fine-tuning to enhance the large model's ability to recognize the dynamic changes in the sheep face.

[0009] Optionally, before the monitoring video data of the target cashmere goats in the breeding area is input into the first goat face recognition module, the method further includes: Obtaining feeding plan data of the breeding area, and predicting the probability of the cashmere goat being in a facial static state according to the feeding plan data and the current time period; wherein the feeding plan data includes feeding time; If the probability is lower than the probability threshold, the monitoring video data of the target cashmere goats in the breeding area is input into the first goat face recognition module; If the probability is higher than the probability threshold, the monitoring video data of the target cashmere goats in the breeding area is input into a second goat face recognition module; wherein the second goat face recognition module includes a convolutional feature extraction model and a goat face recognition model, and the goat face recognition model is constructed based on a classification algorithm.

[0010] Optionally, the monitoring video data of the target cashmere goats in the breeding area is input into the first goat face recognition module, and the first goat face recognition module outputs a first goat face recognition result and a second goat face recognition result, including: Extracting an image sequence containing only the face area of ​​the target cashmere goat from the monitoring video data of the target cashmere goat in the breeding area, and inputting the image sequence into the first goat face recognition module; The first sheep face recognition module extracts the static sheep face data and the dynamic sheep face data from the image sequence, and obtains the first sheep face recognition result based on the analysis of the static sheep face data; The second sheep face recognition result is obtained based on the first sheep face recognition result and the dynamic sheep face data analysis.

[0011] Optionally, obtaining a sheep face recognition result based on the fusion of the first sheep face recognition result and the second sheep face recognition result includes: Determine the weighted weights of the first sheep face recognition result and the second sheep face recognition result respectively; wherein the weighted weight of the first sheep face recognition result is lower than the weighted weight of the second sheep face recognition result; The first sheep face recognition result and the second sheep face recognition result are fused and calculated based on the weighted weight to obtain the sheep face recognition result.

[0012] Optionally, tracking the target cashmere goat according to the sheep face recognition result includes: Acquire tracking project information, and add an identity tag and an additional tag to the target cashmere goat according to the tracking project information.

[0013] The present invention also provides a cashmere goat identification and tracking system based on a large model, the system comprising a processor and a memory, the memory storing executable program code; the processor calls the executable program code stored in the memory to implement the following steps: Collect sheep face data, and use the sheep face data to fine-tune the large model to obtain a first sheep face recognition module; wherein the sheep face data includes static sheep face data and dynamic sheep face data; Inputting the monitoring video data of the target cashmere goats in the breeding area into the first goat face recognition module, the first goat face recognition module outputs a first goat face recognition result and a second goat face recognition result; wherein the first goat face recognition result is obtained based on the analysis of the static goat face data, and the second goat face recognition result is obtained based on the analysis of the dynamic goat face data; Obtaining a sheep face recognition result based on the fusion of the first sheep face recognition result and the second sheep face recognition result; The target cashmere goat is tracked according to the sheep face recognition result.

[0014] Optionally, the collecting of sheep face data, and using the sheep face data to fine-tune the large model to obtain a first sheep face recognition module, comprises: When the face of each cashmere goat in the breeding area is in a static state, the goat face images under multiple angles and multiple lighting conditions are photographed, and the static goat face data is extracted from the goat face images; When each cashmere goat in the breeding area is in a facial activity state, a facial action video is shot, and the dynamic goat face data is extracted from the facial action video; The static sheep face data is used to perform preliminary fine-tuning on the large model to learn the basic features of the sheep face; and the dynamic sheep face data is used to further fine-tune the large model after preliminary fine-tuning to enhance the large model's ability to recognize the dynamic changes in the sheep face.

[0015] Optionally, before the monitoring video data of the target cashmere goats in the breeding area is input into the first goat face recognition module, the method further includes: Obtaining feeding plan data of the breeding area, and predicting the probability of the cashmere goat being in a facial static state according to the feeding plan data and the current time period; wherein the feeding plan data includes feeding time; If the probability is lower than the probability threshold, the monitoring video data of the target cashmere goats in the breeding area is input into the first goat face recognition module; If the probability is higher than the probability threshold, the monitoring video data of the target cashmere goats in the breeding area is input into a second goat face recognition module; wherein the second goat face recognition module includes a convolutional feature extraction model and a goat face recognition model, and the goat face recognition model is constructed based on a classification algorithm.

[0016] Optionally, the monitoring video data of the target cashmere goats in the breeding area is input into the first goat face recognition module, and the first goat face recognition module outputs a first goat face recognition result and a second goat face recognition result, including: Extracting an image sequence containing only the face area of ​​the target cashmere goat from the monitoring video data of the target cashmere goat in the breeding area, and inputting the image sequence into the first goat face recognition module; The first sheep face recognition module extracts the static sheep face data and the dynamic sheep face data from the image sequence, and obtains the first sheep face recognition result based on the analysis of the static sheep face data; The second sheep face recognition result is obtained based on the first sheep face recognition result and the dynamic sheep face data analysis.

[0017] Optionally, obtaining a sheep face recognition result based on the fusion of the first sheep face recognition result and the second sheep face recognition result includes: Determine the weighted weights of the first sheep face recognition result and the second sheep face recognition result respectively; wherein the weighted weight of the first sheep face recognition result is lower than the weighted weight of the second sheep face recognition result; The first sheep face recognition result and the second sheep face recognition result are fused and calculated based on the weighted weight to obtain the sheep face recognition result.

[0018] Optionally, tracking the target cashmere goat according to the sheep face recognition result includes: Acquire tracking project information, and add an identity tag and an additional tag to the target cashmere goat according to the tracking project information.

[0019] The present invention also provides an electronic device, comprising: a memory storing executable program code; a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute any of the methods described above.

[0020] The present invention also provides a computer storage medium, on which a computer program is stored. When the computer program is executed by a processor, any of the above methods is executed.

[0021] The present invention also provides a computer program product, which includes a computer program stored in a computer storage medium, and when the computer program is executed by a processor of an electronic device, it implements any of the methods described above.

[0022] The solution of the present invention simultaneously identifies the cashmere goat based on static information and dynamic information of its face, which can significantly improve the accuracy of cashmere goat identification, and further improve the accuracy of a series of tracking and processing applications on the cashmere goat based on the identity identification information. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 It is a schematic diagram of a cashmere goat identification and tracking method based on a large model disclosed in an embodiment of the present invention.

[0025] Figure 2 The present invention is a schematic diagram of a cashmere goat identification and tracking system based on a large model disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0027] See also Figure 1 The embodiment of the present invention discloses a cashmere goat identification and tracking method based on a large model, the method comprising the following steps: Collect sheep face data, and use the sheep face data to fine-tune the large model to obtain a first sheep face recognition module; wherein the sheep face data includes static sheep face data and dynamic sheep face data; Inputting the monitoring video data of the target cashmere goats in the breeding area into the first goat face recognition module, the first goat face recognition module outputs a first goat face recognition result and a second goat face recognition result; wherein the first goat face recognition result is obtained based on the analysis of the static goat face data, and the second goat face recognition result is obtained based on the analysis of the dynamic goat face data; Obtaining a sheep face recognition result based on the fusion of the first sheep face recognition result and the second sheep face recognition result; The target cashmere goat is tracked according to the sheep face recognition result.

[0028] In the above scheme, the present invention first collects static sheep face data and dynamic sheep face data of cashmere goats to construct training data, uses the training data to fine-tune the general large model, and obtains the first sheep face recognition module. Among them, the general large model specifies OpenAI's GPT series model, Google's BERT model, Microsoft's Azure AI platform, etc. Then, a surveillance camera is arranged in the breeding area, and the surveillance video data corresponding to the target cashmere goat to be identified and tracked is retrieved from the surveillance video shot by the surveillance camera, and the first sheep face recognition module analyzes it. The analysis includes two parts, namely, the first sheep face recognition result about its identity is obtained based on the static sheep face data of the target cashmere goat, and the second sheep face recognition result about its identity is obtained based on the dynamic sheep face data of the target cashmere goat. Then, the fusion algorithm is used to fuse the first sheep face recognition result and the second sheep face recognition result, and then the final sheep face recognition result is obtained. Finally, the target cashmere goat can be tracked according to the sheep face recognition result obtained by identification, and the tracking is used in various application scenarios such as farm management, disease monitoring, and individual tracking.

[0029] The solution of the present invention simultaneously identifies the cashmere goat based on static information and dynamic information of its face, which can significantly improve the accuracy of cashmere goat identification, and further improve the accuracy of a series of tracking and processing applications on the cashmere goat based on the identity identification information.

[0030] Optionally, the collecting of sheep face data, and using the sheep face data to fine-tune the large model to obtain a first sheep face recognition module, comprises: When the face of each cashmere goat in the breeding area is in a static state, the goat face images under multiple angles and multiple lighting conditions are photographed, and the static goat face data is extracted from the goat face images; When each cashmere goat in the breeding area is in a facial activity state, a facial action video is shot, and the dynamic goat face data is extracted from the facial action video; The static sheep face data is used to perform preliminary fine-tuning on the large model to learn the basic features of the sheep face; and the dynamic sheep face data is used to further fine-tune the large model after preliminary fine-tuning to enhance the large model's ability to recognize the dynamic changes in the sheep face.

[0031] In this embodiment, the surveillance cameras arranged in the breeding area can distinguish whether the face of each cashmere goat is in a static state or an active state, wherein the active state is specified as the cashmere goat being in the state of opening its mouth, yawning, eating, ruminating, etc. Then, the face images of the cashmere goat under multiple angles and multiple lighting conditions when the face is in a static state and the face action videos when the face is in an active state are respectively captured, and static face data and dynamic face data of the cashmere goat are respectively extracted.

[0032] At the same time, when fine-tuning the large model, it is first fine-tuned using static sheep face data so that the large model can learn and master the basic features of the sheep face (such as facial features distribution features, fur color features, facial contour features, etc.); then, the large model is further fine-tuned using dynamic sheep face data. At this point, the large model can enhance its ability to recognize the dynamic changes of the sheep face on the basis of accurately identifying the basic features of the sheep face.

[0033] Optionally, before the monitoring video data of the target cashmere goats in the breeding area is input into the first goat face recognition module, the method further includes: Obtaining feeding plan data of the breeding area, and predicting the probability of the cashmere goat being in a facial static state according to the feeding plan data and the current time period; wherein the feeding plan data includes feeding time; If the probability is lower than the probability threshold, the monitoring video data of the target cashmere goats in the breeding area is input into the first goat face recognition module; If the probability is higher than the probability threshold, the monitoring video data of the target cashmere goats in the breeding area is input into a second goat face recognition module; wherein the second goat face recognition module includes a convolutional feature extraction model and a goat face recognition model, and the goat face recognition model is constructed based on a classification algorithm.

[0034] In this embodiment, the present invention constructs two sheep face recognition modules, namely the first sheep face recognition module and the second sheep face recognition module, wherein the first sheep face recognition module is constructed based on a large model, and the second sheep face recognition module is constructed based on a conventional classification algorithm. Compared with the first sheep face recognition module, the second sheep face recognition module has a smaller model volume and a faster measured analysis rate, but the accuracy is slightly lower. In this regard, the present invention first predicts the probability of the cashmere goat being in a static facial state based on the feeding plan data of the breeding area and the current time period. For example, the current time period is far from the feeding time, indicating that the probability of the cashmere goat eating in the current time period is small, so that the probability of its mouth being in an active state is small, that is, the probability of its being in a static facial state at this time is small; on the contrary, the cashmere goat may be in an eating state, and its mouth is likely to be active, that is, the probability of its being in a static facial state at this time is large.

[0035] Therefore, when it is determined that the probability of the cashmere goat being in a static face state is lower than the probability threshold, the surveillance video data of the target cashmere goat shot is directly input into the first goat face recognition module, and the first goat face recognition module obtains its identity information based on the static goat face data and dynamic goat face data of the target cashmere goat. When it is determined that the probability of the cashmere goat being in a static face state is higher than the probability threshold, the second goat face recognition module is used to analyze the surveillance video data of the target cashmere goat in the breeding area, and the analysis is only based on the static goat face data. By selecting the above recognition strategy, the recognition rate can be appropriately improved while ensuring the recognition accuracy.

[0036] The convolution feature extraction model first extracts static sheep face data from the surveillance video data of the target cashmere goat, and then the sheep face recognition model classifies the extracted static sheep face data to obtain the identity of the target cashmere goat, i.e., the sheep face recognition result. The above classification algorithms include but are not limited to support vector machines, decision trees, neural network algorithms, etc.

[0037] Optionally, the monitoring video data of the target cashmere goats in the breeding area is input into the first goat face recognition module, and the first goat face recognition module outputs a first goat face recognition result and a second goat face recognition result, including: Extracting an image sequence containing only the face area of ​​the target cashmere goat from the monitoring video data of the target cashmere goat in the breeding area, and inputting the image sequence into the first goat face recognition module; The first sheep face recognition module extracts the static sheep face data and the dynamic sheep face data from the image sequence, and obtains the first sheep face recognition result based on the analysis of the static sheep face data; The second sheep face recognition result is obtained based on the first sheep face recognition result and the dynamic sheep face data analysis.

[0038] In this embodiment, the first sheep face recognition module should not directly analyze and process the surveillance video data of the target cashmere goat, because the surveillance video of the target cashmere goat will contain its non-facial features, as well as the facial features and non-facial features of other cashmere goats, which will interfere with the identification of the target cashmere goat. In this regard, the present invention is configured to firstly intercept an image sequence containing only the facial area of ​​the target cashmere goat from the surveillance video data of the target cashmere goat, and then the first sheep face recognition module obtains the image sequence for identification processing.

[0039] The first sheep face recognition module first extracts static sheep face data and dynamic sheep face data of the target cashmere goat from the image sequence, and obtains the first sheep face recognition result based on the static sheep face data; and then obtains the second sheep face recognition result based on the first sheep face recognition result and the dynamic sheep face data. Among them, the second identification of the target cashmere goat is performed based on the first sheep face recognition result and the dynamic sheep face data at the same time. Compared with the method of performing identification based on the dynamic sheep face data only, the first sheep face recognition result obtained based on the static sheep face data can be used as a reference, so that the accuracy of the second sheep face recognition result extracted is significantly improved.

[0040] Optionally, obtaining a sheep face recognition result based on the fusion of the first sheep face recognition result and the second sheep face recognition result includes: Determine the weighted weights of the first sheep face recognition result and the second sheep face recognition result respectively; wherein the weighted weight of the first sheep face recognition result is lower than the weighted weight of the second sheep face recognition result; The first sheep face recognition result and the second sheep face recognition result are fused and calculated based on the weighted weight to obtain the sheep face recognition result.

[0041] In this embodiment, since the second sheep face recognition result is essentially obtained based on the analysis of static sheep face data and dynamic sheep face data when the face of the target cashmere goat is in motion, in theory, its confidence level is higher than that of the first sheep face recognition result. Therefore, the present invention sets the weighted weight of the first sheep face recognition result to be lower than that of the second sheep face recognition result, for example, the weighted weight of the first sheep face recognition result is set to 0.4, and the weighted weight of the second sheep face recognition result is set to 0.6.

[0042] Optionally, tracking the target cashmere goat according to the sheep face recognition result includes: Acquire tracking project information, and add an identity tag and an additional tag to the target cashmere goat according to the tracking project information.

[0043] In this embodiment, the tracking project information of the target cashmere goat can be specified according to different needs, such as farm management, disease monitoring, individual tracking, etc. Therefore, after identifying its identity, the identity tag is attached to the target cashmere goat. At the same time, the additional tag is determined according to the tracking project information of the target cashmere goat. For example, when the tracking project information is farm management, the additional tag can be green, and when the farm management is disease monitoring, the additional tag can be red. The above is only for example description and is not intended to limit the scope of protection of the present invention.

[0044] See also Figure 2 As shown, an embodiment of the present invention further provides a cashmere goat identification and tracking system based on a large model, the system comprising a processor and a memory, the memory storing an executable program code; the processor calls the executable program code stored in the memory to implement the following steps: Collect sheep face data, and use the sheep face data to fine-tune the large model to obtain a first sheep face recognition module; wherein the sheep face data includes static sheep face data and dynamic sheep face data; Inputting the monitoring video data of the target cashmere goats in the breeding area into the first goat face recognition module, the first goat face recognition module outputs a first goat face recognition result and a second goat face recognition result; wherein the first goat face recognition result is obtained based on the analysis of the static goat face data, and the second goat face recognition result is obtained based on the analysis of the dynamic goat face data; Obtaining a sheep face recognition result based on the fusion of the first sheep face recognition result and the second sheep face recognition result; The target cashmere goat is tracked according to the sheep face recognition result.

[0045] Optionally, the collecting of sheep face data, and using the sheep face data to fine-tune the large model to obtain a first sheep face recognition module, comprises: When the face of each cashmere goat in the breeding area is in a static state, the goat face images under multiple angles and multiple lighting conditions are photographed, and the static goat face data is extracted from the goat face images; When each cashmere goat in the breeding area is in a facial activity state, a facial action video is shot, and the dynamic goat face data is extracted from the facial action video; The static sheep face data is used to perform preliminary fine-tuning on the large model to learn the basic features of the sheep face; and the dynamic sheep face data is used to further fine-tune the large model after preliminary fine-tuning to enhance the large model's ability to recognize the dynamic changes in the sheep face.

[0046] Optionally, before the monitoring video data of the target cashmere goats in the breeding area is input into the first goat face recognition module, the method further includes: Obtaining feeding plan data of the breeding area, and predicting the probability of the cashmere goat being in a facial static state according to the feeding plan data and the current time period; wherein the feeding plan data includes feeding time; If the probability is lower than the probability threshold, the monitoring video data of the target cashmere goats in the breeding area is input into the first goat face recognition module; If the probability is higher than the probability threshold, the monitoring video data of the target cashmere goats in the breeding area is input into a second goat face recognition module; wherein the second goat face recognition module includes a convolutional feature extraction model and a goat face recognition model, and the goat face recognition model is constructed based on a classification algorithm.

[0047] Optionally, the monitoring video data of the target cashmere goats in the breeding area is input into the first goat face recognition module, and the first goat face recognition module outputs a first goat face recognition result and a second goat face recognition result, including: Extracting an image sequence containing only the face area of ​​the target cashmere goat from the monitoring video data of the target cashmere goat in the breeding area, and inputting the image sequence into the first goat face recognition module; The first sheep face recognition module extracts the static sheep face data and the dynamic sheep face data from the image sequence, and obtains the first sheep face recognition result based on the analysis of the static sheep face data; The second sheep face recognition result is obtained based on the first sheep face recognition result and the dynamic sheep face data analysis.

[0048] Optionally, obtaining a sheep face recognition result based on the fusion of the first sheep face recognition result and the second sheep face recognition result includes: Determine the weighted weights of the first sheep face recognition result and the second sheep face recognition result respectively; wherein the weighted weight of the first sheep face recognition result is lower than the weighted weight of the second sheep face recognition result; The first sheep face recognition result and the second sheep face recognition result are fused and calculated based on the weighted weight to obtain the sheep face recognition result.

[0049] Optionally, tracking the target cashmere goat according to the sheep face recognition result includes: Acquire tracking project information, and add an identity tag and an additional tag to the target cashmere goat according to the tracking project information.

[0050] An embodiment of the present invention further provides an electronic device, comprising: a memory storing executable program code; a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the method described in any of the above embodiments.

[0051] An embodiment of the present invention further provides a computer storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method described in any of the above embodiments is executed.

[0052] An embodiment of the present invention further provides a computer program product, which includes a computer program stored in a computer storage medium, and when the computer program is executed by a processor of an electronic device, the method described in any of the above embodiments is implemented.

[0053] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (apparatus), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0054] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0055] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0056] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A cashmere goat identification and tracking method based on a large model, characterized in that: The method comprises the following steps: Collect sheep face data, and use the sheep face data to fine-tune the large model to obtain a first sheep face recognition module; wherein the sheep face data includes static sheep face data and dynamic sheep face data; Inputting the monitoring video data of the target cashmere goats in the breeding area into the first goat face recognition module, the first goat face recognition module outputs a first goat face recognition result and a second goat face recognition result; wherein the first goat face recognition result is obtained based on the analysis of the static goat face data, and the second goat face recognition result is obtained based on the analysis of the dynamic goat face data; Obtaining a sheep face recognition result based on the fusion of the first sheep face recognition result and the second sheep face recognition result; The target cashmere goat is tracked according to the sheep face recognition result.

2. The cashmere goat identification and tracking method based on a large model according to claim 1, characterized in that: The step of collecting sheep face data and using the sheep face data to fine-tune the large model to obtain a first sheep face recognition module includes: When the face of each cashmere goat in the breeding area is in a static state, the goat face images under multiple angles and multiple lighting conditions are photographed, and the static goat face data is extracted from the goat face images; When each cashmere goat in the breeding area is in a facial activity state, a facial action video is shot, and the dynamic goat face data is extracted from the facial action video; The static sheep face data is used to perform preliminary fine-tuning on the large model to learn the basic features of the sheep face; and the dynamic sheep face data is used to further fine-tune the large model after preliminary fine-tuning to enhance the large model's ability to recognize the dynamic changes in the sheep face.

3. A cashmere goat identification and tracking method based on a large model according to claim 2, characterized in that: Before inputting the monitoring video data of the target cashmere goats in the breeding area into the first goat face recognition module, the method further includes: Obtaining feeding plan data of the breeding area, and predicting the probability of the cashmere goat being in a facial static state according to the feeding plan data and the current time period; wherein the feeding plan data includes feeding time; If the probability is lower than the probability threshold, the monitoring video data of the target cashmere goats in the breeding area is input into the first goat face recognition module; If the probability is higher than the probability threshold, the monitoring video data of the target cashmere goats in the breeding area is input into a second goat face recognition module; wherein the second goat face recognition module includes a convolutional feature extraction model and a goat face recognition model, and the goat face recognition model is constructed based on a classification algorithm.

4. The cashmere goat identification and tracking method based on a large model according to claim 3 is characterized in that: The monitoring video data of the target cashmere goats in the breeding area is input into the first goat face recognition module, and the first goat face recognition module outputs a first goat face recognition result and a second goat face recognition result, including: Extracting an image sequence containing only the face area of ​​the target cashmere goat from the monitoring video data of the target cashmere goat in the breeding area, and inputting the image sequence into the first goat face recognition module; The first sheep face recognition module extracts the static sheep face data and the dynamic sheep face data from the image sequence, and obtains the first sheep face recognition result based on the analysis of the static sheep face data; The second sheep face recognition result is obtained based on the first sheep face recognition result and the dynamic sheep face data analysis.

5. The cashmere goat identification and tracking method based on a large model according to claim 4 is characterized in that: Obtaining a sheep face recognition result based on the fusion of the first sheep face recognition result and the second sheep face recognition result, including: Determine the weighted weights of the first sheep face recognition result and the second sheep face recognition result respectively; wherein the weighted weight of the first sheep face recognition result is lower than the weighted weight of the second sheep face recognition result; The first sheep face recognition result and the second sheep face recognition result are fused and calculated based on the weighted weight to obtain the sheep face recognition result.

6. The cashmere goat identification and tracking method based on a large model according to claim 5 is characterized in that: Tracking the target cashmere goat according to the sheep face recognition result includes: Acquire tracking project information, and add an identity tag and an additional tag to the target cashmere goat according to the tracking project information.

7. A cashmere goat identification and tracking system based on a large model, the system comprising a processor and a memory, the memory storing executable program code; characterized in that: The processor calls the executable program code stored in the memory to implement the following steps: Collect sheep face data, and use the sheep face data to fine-tune the large model to obtain a first sheep face recognition module; wherein the sheep face data includes static sheep face data and dynamic sheep face data; Inputting the monitoring video data of the target cashmere goats in the breeding area into the first goat face recognition module, the first goat face recognition module outputs a first goat face recognition result and a second goat face recognition result; wherein the first goat face recognition result is obtained based on the analysis of the static goat face data, and the second goat face recognition result is obtained based on the analysis of the dynamic goat face data; Obtaining a sheep face recognition result based on the fusion of the first sheep face recognition result and the second sheep face recognition result; The target cashmere goat is tracked according to the sheep face recognition result.

8. An electronic device comprising: A memory storing executable program code; A processor coupled to the memory; characterized in that: the processor calls the executable program code stored in the memory to execute the method according to any one of claims 1-6.

9. A computer storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is executed.

10. A computer program product, comprising a computer program stored in a computer storage medium, characterized in that: When the computer program is executed by a processor of an electronic device, the method according to any one of claims 1 to 6 is implemented.

Images