Animal identity recognition authentication and re-recognition system based on multi-machine collaborative visual perception

Abstract

The invention discloses an animal identity recognition authentication and re-recognition system based on multi-machine collaborative visual perception. The animal identity recognition authentication and re-recognition system comprises an information acquisition device, a data processing module and an intelligent platform display device, the information acquisition device is used for carrying out all-around information acquisition on animals in a farm, and the information acquisition comprises shooting pictures, videos and related environment information of the animals; the information acquisition device comprises an inspection track, a fixed camera, an inspection camera and a portable camera device, the inspection camera is arranged on the inspection track, and the data processing module adopts a pre-trained deep learning model to carry out target detection, semantic segmentation and re-identification technology processing to acquire animal information; the intelligent platform display device receives the data from the data processing module; according to the scheme, equipment investment is reduced, limitation of distance and environment is avoided, recognition precision is improved, non-contact authentication is beneficial to reduction of animal stress response, and healthy growth and production performance exertion of animals are facilitated.

Description

Technical Field

[0001] The present invention belongs to the technical field of animal husbandry visual recognition, and in particular relates to an animal identity recognition authentication and re-recognition system based on multi-machine collaborative visual perception. Background Art

[0002] Re-ID technology, also known as re-recognition technology, is a technology that uses computer vision technology or related algorithms to identify specific targets (such as pedestrians, vehicles, animals, etc.) from images or video sequences; re-ID technology aims to solve the problem of cross-camera, cross-time, and cross-view target recognition. It extracts the characteristic information of the target object (such as appearance, behavior pattern, etc.) and compares and matches it in different images or video frames, thereby achieving cross-camera tracking and recognition of the target.

[0003] Lean management of livestock farms requires accurate identification of each animal, and the uniqueness of each animal is determined by the unique identifier assigned (abbreviated as ID). Traditional animal identification methods usually use physical devices such as ear tags, collars, and leg rings to identify the animal. The physical device stores the assigned unique number as the animal's ID, which is read out by a dedicated reading device. For example, in a cattle farm, when specific activities occur, such as feeding, milking, breeding, vaccination, etc., the ID needs to be obtained to record the activity information of the individual "cow".

[0004] However, this traditional animal identification method requires the installation of special equipment on the animals and the deployment of special reading devices at specific locations in the animal's activity area. The larger the number of animals, the more equipment needs to be deployed, resulting in high management costs for the company. In addition, the ID storage device and reading device can only work at a short distance and cannot provide IDs throughout the animal's entire activity space. The reading position is limited and the reading accuracy is affected by environmental interference.

[0005] The resulting loose management of animals and inaccurate management of individual animals may lead to certain dangers erupting from individuals, ultimately leading to overall adverse reactions, etc. Therefore, this solution provides an animal identity recognition and re-identification system based on multi-machine collaborative visual perception to solve the problems raised by the above background technology. Summary of the invention

[0006] In response to the problems raised by the above background technology, the purpose of the present invention is to provide an animal identity recognition authentication and re-identification system based on multi-machine collaborative visual perception, which reduces equipment investment, improves recognition accuracy and is not limited by distance and environment. Automatic data aggregation and statistics saves manual analysis time. At the same time, contactless authentication is beneficial to reducing animal stress response and is conducive to their healthy growth and production performance.

[0007] In order to achieve the above technical objectives, the technical solution adopted by the present invention is as follows:

[0008] An animal identification authentication and re-identification system based on multi-machine collaborative visual perception, including an information collection device, a data processing module and an intelligent platform display device;

[0009] The information collection device is used to collect all-round information about the animals in the farm, and the information collection includes taking pictures, videos and related environmental information of the animals; the information collection device includes a patrol track, a fixed camera, a patrol camera and a portable camera device, the patrol camera is arranged on the patrol track, and the patrol track is arranged around the farm; the information collection device transmits the collected information to the data processing module;

[0010] The data processing module includes a server, the server acquires data sent by the information collection device, and the data processing module uses a pre-trained deep learning model to perform target detection, semantic segmentation and re-identification technology to process the collected animal information and generate corresponding recognition and association results; the recognition and association results are sent to the smart platform display device; the recognition and association results include the identity, location and status information of the animal;

[0011] The smart platform display device receives data from the data processing module and is also used to display the environmental information of the farm, the identity, location and status information of the animals.

[0012] It is further defined that the information collection device sends the collected image or video data to the data processing module through the communication interface.

[0013] Further defining, comprising the following steps:

[0014] Step 1: The user selects the corresponding information collection plan based on the camera type and location in the farm, and collects animal information;

[0015] Step 2: The animal information collected in step 1 is transmitted to the data processing module through edge cloud computing technology;

[0016] Step 3: The data processing module uses target detection, semantic segmentation and re-identification technology to process the collected animal information.

[0017] Generate corresponding identification and association results; the animal information includes animal pictures or animal videos;

[0018] Step 4: Store the results generated in step 3 into the database for subsequent analysis and application;

[0019] Step 5: Push the results generated in step 3 to the front-end display for users to view and make decisions.

[0020] It is further defined that the step 1 specifically comprises: using a fixed camera, a patrol camera or a portable camera device to collect all-round information on the animals in the farm;

[0021] The step 2 specifically includes: transmitting the collected animal information to a data processing module, and using edge cloud computing technology for data processing and storage;

[0022] The step three is specifically as follows: performing target detection on the collected animal pictures or animal videos, and using computer vision technology to identify the animals in the pictures; then performing semantic segmentation on the detected animals, separating them from the background, so as to extract more accurate features; and finally using re-identification technology to identify the same animal collected in different time periods, and associate them to ensure accurate authentication of the animal's identity;

[0023] The step 4 specifically includes: storing the data processed in step 3 into a database, wherein the data stored in the database includes the attribute information, identification results and environmental data of the animal;

[0024] The step five is specifically as follows: the identification and association results generated in step three are pushed to the front-end display, and the user can view the identity authentication and re-identification status of the animal and make corresponding decisions based on the user's actual needs.

[0025] It is further defined that the deep learning model includes a small sample data model, and the small sample data model expansion technology solves the problem of small amount of training data due to uncontrollable animal behavior, and the steps include:

[0026] A small number of samples are extracted for attribute feature extraction, coupled feature space is performed, and attribute vectors are output; then a small number of samples are randomly amplified by flipping, rotating or cropping to obtain the amplified real samples; then the attribute features are embedded in the network, an adversarial network model is generated through the generator, and the key attributes of the samples are extracted through the semantically constrained adversarial network, and finally samples with diversified attributes are output.

[0027] It is further defined that the deep learning model also includes a data sample adaptive expansion training technology, and the adaptive expansion training technology performs adaptive expansion training on the model based on an active learning strategy of entropy bagging query, including the following steps:

[0028] By setting an unbalanced data set, the unbalanced data set includes a training set and a test set, multiple classifiers are trained in the training set, and after the training is completed, each sample is input into the classifier to generate a label, and then the most representative sample is selected for manual labeling to obtain a labeled sample and use it to establish a model to achieve adaptive expansion training;

[0029] The expression used by the data sample adaptive expansion training technology is formula 1:

[0030]

[0031] Where H is the sample x i The entropy value, p(y * =w|x i ) represents the sample x i The probability of being predicted by the model as category w, y * is the prediction result of the model, N i is the total number of categories.

[0032] It is further specified that the deep learning model also includes an adaptive learning strategy, which learns new unlabeled images, optimizes the knowledge graph, and shortens the model training cycle.

[0033] It is further defined that the re-identification technology performs identity recognition on individual animals, and the re-identification network simultaneously extracts the global features and local features of the animal, and extracts the gait features of the animal based on two adjacent frames of images to comprehensively recognize the identity information of the animal.

[0034] Beneficial effects of the present invention:

[0035] 1. This solution provides an animal identification authentication and re-identification system based on multi-machine collaborative visual perception, including the following steps: using a camera to collect all-round information on animals in a farm, including taking pictures of animals and related environmental information; performing target detection on the collected animal pictures or videos, using computer vision technology to identify the animals in the pictures, and performing semantic segmentation on the target detected animals to separate them from the background, using re-identification technology to identify the same animal collected in different time periods and associate them, which can be achieved through a pre-trained deep learning model or method; storing the processed data in a database for subsequent analysis and application. The database can store the attribute information, identification results and related environmental data of the animals. The processed data is stored in a database for subsequent analysis and application. The database can store the attribute information, identification results and related environmental data of the animals. The generated ranking, recommendation, report and other results are pushed to the front-end display, which can be an intelligent management platform or other related applications. Users can view the situation of the animals, understand the test results, and make corresponding decisions according to actual needs.

[0036] 2. In terms of cost, by setting up information collection devices, data processing modules based on visual perception and inspection tracks to collect data from farms, the equipment investment is reduced compared to existing designs. In terms of recognition accuracy, animal information is collected in all directions, and computer vision technology is used for target detection, semantic segmentation and re-identification, which is not restricted by distance and environment and has higher accuracy. In animal management, the conditions of individual animals at different times are accurately identified and associated, high-yield individuals are automatically classified, and management strategies are optimized. In terms of efficiency, 24-hour real-time observation and automatic data summary and statistics save manual analysis time. At the same time, contactless authentication reduces animal stress response, which is beneficial to their healthy growth and production performance. In summary, the present invention effectively solves the technical problems that traditional methods are relatively loose in animal management and are not accurate enough in the management of individual animals, and at the same time brings innovation to animal breeding management. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The present invention can be further illustrated by the non-limiting examples given in the accompanying drawings;

[0038] Figure 1 A schematic diagram of the deployment of an embodiment of an animal identity recognition, authentication and re-identification system based on multi-machine collaborative visual perception according to the present invention;

[0039] Figure 2 It is a schematic diagram of a small sample data model expansion of an embodiment of an animal identity recognition authentication and re-recognition system based on multi-machine collaborative visual perception of the present invention;

[0040] Figure 3 It is a schematic diagram of adaptive expansion training of an embodiment of an animal identity recognition authentication and re-recognition system based on multi-machine collaborative visual perception of the present invention;

[0041] Figure 4 A schematic diagram of adaptive transfer learning of an embodiment of an animal identity recognition, authentication and re-identification system based on multi-machine collaborative visual perception according to the present invention;

[0042] Figure 5 A schematic diagram of target re-identification of an embodiment of an animal identity recognition authentication and re-identification system based on multi-machine collaborative visual perception according to the present invention;

[0043] Figure 6 It is a flow chart of an embodiment of an animal identity recognition authentication and re-identification system based on multi-machine collaborative visual perception of the present invention;

[0044] Figure 7 This is a display interface diagram of a smart platform display device of an embodiment of an animal identity recognition, authentication and re-identification system based on multi-machine collaborative visual perception of the present invention. DETAILED DESCRIPTION

[0045] In order to enable those skilled in the art to better understand the present invention, the technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments. The technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings 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 the field without creative work are within the scope of protection of the present invention.

[0046] It should be noted that all directional indications in the embodiments of the present invention (such as up, down, left, right, front, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0047] In addition, the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but it must be based on the ability of ordinary technicians in the field to implement. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0048] like Figure 1-Figure 6 As shown, an animal identification authentication and re-identification system based on multi-machine collaborative visual perception of the present invention includes an information collection device, a data processing module and a smart platform display device;

[0049] The information collection device is used to collect all-round information about the animals in the farm, and the information collection includes taking pictures, videos and related environmental information of the animals; the information collection device includes a patrol track, a fixed camera, a patrol camera and a portable camera device, the patrol camera is arranged on the patrol track, and the patrol track is arranged around the farm; the information collection device transmits the collected information to the data processing module;

[0050] Specifically, the information collection device of the present scheme uses a high-resolution camera to collect images or videos of animals in a farm and record relevant environmental information; the camera can be fixedly installed in the farm or collected by staff in the form of a mobile device; the staff can use a portable camera to take photos or videos of animals in the farm and upload them to a server for further processing and analysis; the information collection device sends the collected image or video data to a data processing module through a communication interface.

[0051] like Figure 6 As shown in the figure, the process of fixed cameras is: read video → target detection → semantic segmentation → re-identification → individual ID → activity change → store in database → ranking → recommendation → report generation → front-end display;

[0052] The process of the inspection camera is: read video → target detection → semantic segmentation → re-identification → individual ID → activity change → store in database → classification → front-end display;

[0053] The data processing module includes a server, the server acquires data sent by the information collection device, and the data processing module uses a pre-trained deep learning model to perform target detection, semantic segmentation and re-identification technology to process the collected animal information and generate corresponding recognition and association results; the recognition and association results are sent to the smart platform display device; the recognition and association results include the identity, location and status information of the animal;

[0054] The smart platform display device receives data from the data processing module and is also used to display the environmental information of the farm, the identity, location and status information of the animals.

[0055] Preferably, the information collection device sends the collected image or video data to the data processing module through a communication interface.

[0056] Preferably, the method comprises the following steps:

[0057] Step 1: The user selects the corresponding information collection plan based on the camera type and location in the farm, and collects animal information;

[0058] Step 2: The animal information collected in step 1 is transmitted to the data processing module through edge cloud computing technology;

[0059] Step 3: The data processing module processes the collected animal information using target detection, semantic segmentation and re-identification technology to generate corresponding identification and association results; the animal information includes animal pictures or animal videos;

[0060] Step 4: Store the results generated in step 3 into the database for subsequent analysis and application;

[0061] Step 5: Push the results generated in step 3 to the front-end display for users to view and make decisions.

[0062] Preferably, the step 1 specifically comprises: using a fixed camera, a patrol camera or a portable camera device to collect all-round information on the animals in the farm;

[0063] The step 2 specifically includes: transmitting the collected animal information to a data processing module, and using edge cloud computing technology for data processing and storage;

[0064] The step three is specifically as follows: performing target detection on the collected animal pictures or animal videos, and using computer vision technology to identify the animals in the pictures; then performing semantic segmentation on the detected animals, separating them from the background, so as to extract more accurate features; and finally using re-identification technology to identify the same animal collected in different time periods, and associate them to ensure accurate authentication of the animal's identity;

[0065] The step 4 specifically includes: storing the data processed in step 3 into a database, wherein the data stored in the database includes the attribute information, identification results and environmental data of the animal;

[0066] The step five is specifically as follows: the identification and association results generated in step three are pushed to the front-end display, and the user can view the identity authentication and re-identification status of the animal and make corresponding decisions based on the user's actual needs.

[0067] like Figure 2 As shown, preferably, the deep learning model includes a small sample data model, and the small sample data model expansion technology solves the problem of small amount of training data due to uncontrollable animal behavior, and the steps include:

[0068] A small number of samples are extracted for attribute feature extraction, coupled feature space is performed, and attribute vectors are output; then a small number of samples are randomly amplified by flipping, rotating or cropping to obtain the amplified real samples; then the attribute features are embedded in the network, an adversarial network model is generated through the generator, and the key attributes of the samples are extracted through the semantically constrained adversarial network, and finally samples with diversified attributes are output.

[0069] like Figure 3 As shown, preferably, the deep learning model also includes a data sample adaptive expansion training technology, and the adaptive expansion training technology performs adaptive expansion training on the model based on an active learning strategy of entropy bagging query, including the following steps:

[0070] By setting an unbalanced data set, the unbalanced data set includes a training set and a test set, multiple classifiers are trained in the training set, and after the training is completed, each sample is input into the classifier to generate a label, and then the most representative sample is selected for manual labeling to obtain a labeled sample and use it to establish a model to achieve adaptive expansion training;

[0071] The expression used by the data sample adaptive expansion training technology is formula 1:

[0072]

[0073] Where H is the sample x i The entropy value, p(y * =w|x i ) represents the sample x i The probability of being predicted by the model as category w, y * is the prediction result of the model, N i is the total number of categories.

[0074] like Figure 4 As shown, preferably, the deep learning model also includes an adaptive learning strategy, which learns new unlabeled images, optimizes the knowledge graph, and shortens the model training cycle.

[0075] like Figure 5 As shown, preferably, the re-identification technology identifies individual animals, and the re-identification network extracts the global and local features of the animal at the same time, and extracts the gait features of the animal based on two adjacent frames of images to comprehensively identify the identity information of the animal. Using contactless identity authentication for animals can reduce the stress response of animals, conduct real-time observation 24 hours a day, automatically summarize and count the data, and automatically classify high-yield individuals in a concise and clear manner, saving the process of manual data analysis.

[0076] The working principle and effect of this solution are:

[0077] This solution provides an animal identity authentication and re-identification system based on multi-machine collaborative visual perception. In terms of cost, by setting up an information collection device, a data processing module based on visual perception, and an inspection track to collect farm data, the equipment investment is reduced compared to the existing design. In terms of recognition accuracy, animal information is collected in all directions, and computer vision technology is used for target detection, semantic segmentation, and re-identification, which is not limited by distance and environment and has higher accuracy. In animal management, the conditions of individual animals at different time periods are accurately identified and associated, high-yield individuals are automatically classified, and management strategies are optimized. In terms of efficiency, 24-hour real-time observation and automatic data summary and statistics save manual analysis time. At the same time, contactless authentication reduces animal stress response, which is beneficial to their healthy growth and production performance. In summary, the present invention effectively solves the technical problems that traditional methods are relatively loose in animal management and not precise in the management of individual animals, and at the same time brings innovation to animal breeding management.

[0078] The display interface of the smart platform display device in this embodiment is as follows Figure 7 As shown, as an example of this embodiment, the example is a display interface of breeding geese, specifically: the classification of breeding geese in the circle and their fixed identities are displayed in real time on the large screen in the middle, and different categories are distinguished by different colors;

[0079] The right side of the display interface is the egg production ranking, which includes the monthly egg production and total egg production statistics page, and displays the ranking and individual images at the same time; the left side of the display interface is the pen information and individual recommendation information. The background algorithm scores and recommends the animals required by the user. The user can make different recommendations according to their needs. For example, the information of one goose is ID: 2, with a score of 100; the information of another goose is ID: 3 with a score of 90;

[0080] The bottom of the display interface is a comprehensive information display, including the total egg production, temperature, humidity, etc. of the pen. The displayed data here can be changed according to user needs to meet the user's management of the pen. The above is only one example of this invention, and the specific example of this invention can be changed according to user needs.

[0081] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the technology may modify or change the above embodiments without violating the spirit and scope of the present invention. Therefore, all equivalent modifications or changes made by those with ordinary knowledge in the technical field without departing from the spirit and technical ideas disclosed by the present invention should still be covered by the claims of the present invention.

Claims

1. An animal identification authentication and re-identification system based on multi-machine collaborative visual perception, characterized by: It includes information collection device, data processing module and smart platform display device; The information collection device is used to collect all-round information about the animals in the farm, and the information collection includes taking pictures, videos and related environmental information of the animals; the information collection device includes a patrol track, a fixed camera, a patrol camera and a portable camera device, the patrol camera is arranged on the patrol track, and the patrol track is arranged around the farm; the information collection device transmits the collected information to the data processing module; The data processing module includes a server, the server acquires data sent by the information collection device, and the data processing module uses a pre-trained deep learning model to perform target detection, semantic segmentation and re-identification technology to process the collected animal information and generate corresponding recognition and association results; the recognition and association results are sent to the smart platform display device; the recognition and association results include the identity, location and status information of the animal; The smart platform display device receives data from the data processing module and is also used to display the environmental information of the farm, the identity, location and status information of the animals.

2. The animal identification authentication and re-identification system based on multi-machine collaborative visual perception according to claim 1, characterized in that: The information collection device sends the collected picture or video data to the data processing module through the communication interface.

3. The animal identification authentication and re-identification system based on multi-machine collaborative visual perception according to claim 1 is characterized in that: The steps include: Step 1: The user selects the corresponding information collection plan based on the camera type and location in the farm, and collects animal information; Step 2: The animal information collected in step 1 is transmitted to the data processing module through edge cloud computing technology; Step 3: The data processing module uses target detection, semantic segmentation and re-identification technology to process the collected animal information. Generate corresponding identification and association results; the animal information includes animal pictures or animal videos; Step 4: Store the results generated in step 3 into the database for subsequent analysis and application; Step 5: Push the results generated in step 3 to the front-end display for users to view and make decisions.

4. The animal identification authentication and re-identification system based on multi-machine collaborative visual perception according to claim 3 is characterized in that: The step 1 specifically includes: using fixed cameras, patrol cameras or portable cameras to collect all-round information about the animals in the farm; The step 2 specifically includes: transmitting the collected animal information to a data processing module, and using edge cloud computing technology for data processing and storage; The step three is specifically as follows: performing target detection on the collected animal pictures or animal videos, and using computer vision technology to identify the animals in the pictures; then performing semantic segmentation on the detected animals, separating them from the background, so as to extract more accurate features; and finally using re-identification technology to identify the same animal collected in different time periods, and associate them to ensure accurate authentication of the animal's identity; The step 4 specifically includes: storing the data processed in step 3 into a database, wherein the data stored in the database includes the attribute information, identification results and environmental data of the animal; The step five is specifically as follows: the identification and association results generated in step three are pushed to the front-end display, and the user can view the identity authentication and re-identification status of the animal and make corresponding decisions based on the user's actual needs.

5. The animal identification authentication and re-identification system based on multi-machine collaborative visual perception according to claim 1 is characterized in that: The deep learning model includes a small sample data model. The small sample data model expansion technology solves the problem of small amount of training data due to uncontrollable animal behavior, and the steps include: A small number of samples are extracted for attribute feature extraction, coupled feature space is performed, and attribute vectors are output; then a small number of samples are randomly amplified by flipping, rotating or cropping to obtain the amplified real samples; then the attribute features are embedded in the network, an adversarial network model is generated through the generator, and the key attributes of the samples are extracted through the semantically constrained adversarial network, and finally samples with diversified attributes are output.

6. The animal identification authentication and re-identification system based on multi-machine collaborative visual perception according to claim 1, characterized in that: The deep learning model also includes a data sample adaptive expansion training technology, which performs adaptive expansion training on the model based on an active learning strategy of entropy bagging query, including the following steps: By setting an unbalanced data set, the unbalanced data set includes a training set and a test set, multiple classifiers are trained in the training set, and after the training is completed, each sample is input into the classifier to generate a label, and then the most representative sample is selected for manual labeling to obtain a labeled sample and use it to establish a model to achieve adaptive expansion training; The expression used by the data sample adaptive expansion training technology is formula 1: Where H is the sample x i The entropy value, p(y * =w|x i ) represents the sample x i The probability of being predicted by the model as category w, y * is the prediction result of the model, N i is the total number of categories.

7. The animal identity recognition and re-identification system based on multi-machine collaborative visual perception according to claim 6, characterized in that: The deep learning model also includes an adaptive learning strategy, which learns new unlabeled images, optimizes the knowledge graph, and shortens the model training cycle.

8. The animal identification authentication and re-identification system based on multi-machine collaborative visual perception according to claim 1, characterized in that: The re-identification technology identifies the individual animal. The re-identification network extracts the global features and local features of the animal at the same time, and extracts the gait features of the animal based on two adjacent frames of images to comprehensively identify the identity information of the animal.

Images