Animal social barrier behavior detection device and method, electronic equipment and medium

Abstract

The invention relates to the technical field of animal behavioristics, in particular to an animal social barrier behavior detection device and method, electronic equipment and a medium. The animal social barrier behavior detection device is composed of an observation box body, an image acquisition device and a data processing unit. The interior of the observation box body is divided into a first sub-chamber and a second sub-chamber which are respectively used for placing experimental animals and contrast animals, so that behavior observation of the two groups of animals in the same environment is ensured. And the multi-view image acquisition device in each sub-chamber captures animal activities from different angles through a plurality of cameras to generate multi-view activity pictures. And the data processing unit is responsible for extracting behavior tracks of the animals from the pictures and analyzing the tracks by using a pre-trained phenotype classification model to detect social barrier behaviors, so that efficient and quantitative analysis of animal behaviors is realized.

Description

Technical Field

[0001] This application relates to the field of animal behavior technology, and in particular to an animal social disorder behavior detection device and method, electronic device, and medium. Background Technology

[0002] Traditional detection methods for social disorder behaviors in animals mainly rely on manual observation or two-dimensional video analysis. Typical paradigms include open field tests, three-box social experiments, social interaction tests, and object recognition tests. These methods semi-quantitatively evaluate mice's social approach and avoidance, ability to explore novel environments, and stereotyped repetitive behaviors by recording the animal's position trajectory in a specific scenario and the frequency and duration of contact with target objects or companions.

[0003] However, such methods have limitations in terms of data acquisition dimensions and behavioral decoding depth. They can only achieve planar trajectory tracking from a single camera perspective, obtaining coarse-grained spatial information about "where the animal is active," but cannot accurately identify the fine semantic meaning of "what behavior is being performed." Especially in the quantification of stereotyped behaviors, due to the lack of continuous capture of three-dimensional movements of key body parts, it is difficult to objectively determine the start time, duration, and evolutionary pattern of abnormal repetitive movements, and it is even more impossible to distinguish specific behavioral subtypes such as grooming, jumping, and spinning. This leads to a large amount of subjective judgment bias in the evaluation results, and the consistency coefficient between different experimenters is usually lower than the benchmark requirements for reproducibility studies. This one-sidedness makes it difficult for the detection data to fully reflect the core phenotypic characteristics of autism model mice, reducing the efficiency of detecting animal social disorder behaviors. Therefore, how to efficiently detect animal behavioral activities to better achieve quantitative analysis of animal behavioral activities has become a major problem that urgently needs to be solved in the field. Summary of the Invention

[0004] This application aims to address at least one of the technical problems existing in the prior art. To this end, this application proposes an animal social behavior detection device and method, electronic device, and medium, which can efficiently detect animal behavior activities to better achieve quantitative analysis of animal behavior activities.

[0005] An animal social disorder behavior detection device according to a first aspect embodiment of this application includes: An observation box, the inner side of which forms an observation cavity, is used to provide space for animal behavior and activity; wherein, the interior of the observation box is divided into a first sub-chamber and a second sub-chamber that are isolated from each other, the first sub-chamber is used to contain experimental animals, and the second sub-chamber is used to simultaneously contain control animals under the same environmental conditions as the first sub-chamber, wherein the experimental animals and the control animals are the same species of animals that exhibit different behavioral behaviors; An image acquisition device includes multiple cameras set from different shooting angles, used to capture the behavior and activities of animals from the multiple shooting angles, each of the cameras corresponding to one of the shooting angles; The image acquisition device installed in the first sub-chamber is a first multi-view image acquisition device, which is used to acquire experimental multi-view activity images of the experimental animal; The image acquisition device installed in the second sub-chamber is a second multi-view image acquisition device, which is used to acquire control multi-view activity images of the control animal; A data processing unit is connected to the first multi-view image acquisition device and the second multi-view image acquisition device. The data processing unit is used to extract the behavioral trajectory of the experimental animal from the experimental multi-view activity screen and extract the behavioral trajectory of the control animal from the control multi-view activity screen. Then, it uses a pre-trained phenotypic classification model to detect social disorder behavior in the experimental animal behavioral trajectory and the control animal behavioral trajectory to obtain animal social disorder behavior detection data.

[0006] The method for detecting animal social disorder behavior according to the second aspect of this application, applied to the animal social disorder behavior detection device described in the first aspect of this application, includes: The experimental animal housed in the first sub-chamber is captured by the first multi-view image acquisition device to obtain multi-view images of the experimental animal's activities. The control animal housed in the second sub-chamber is captured by a second multi-view image acquisition device to obtain a control multi-view activity image of the control animal. Extract the behavioral trajectories of the experimental animals from the multi-view activity footage of the experiment; Extract the behavioral trajectories of the control animals from the aforementioned multi-view activity footage; Social disorder behavior detection data of the experimental animals and control animals was obtained by using a pre-trained phenotypic classification model to detect the behavioral trajectories of the experimental animals and control animals.

[0007] According to some embodiments of this application, before the step of detecting social disorder behavior by using a pre-trained phenotypic classification model to analyze the behavioral trajectories of the experimental animals and control animals to obtain animal social disorder behavior detection data, the method further includes pre-training the phenotypic classification model, specifically including: Determine the original classification and detection model; Obtain a behavior detection training dataset; wherein, the behavior detection training dataset includes multiple behavior phenotype sample trajectories and behavior phenotype annotation data corresponding to each behavior phenotype sample trajectory; Based on the sample trajectories of each behavioral phenotype and the behavioral phenotype annotation data corresponding to each sample trajectories, the classification detection model is trained under supervision to obtain the pre-trained phenotype classification model.

[0008] According to some embodiments of this application, obtaining the behavior detection training dataset includes: Obtain multiple behavioral phenotype sample trajectories; wherein each behavioral phenotype sample trajectory corresponds to a sample behavioral phenotype of a sample animal; Two-dimensional body tracking was performed on the trajectory of each behavioral phenotype sample to obtain several sample body tracking data. Three-dimensional behavior reconstruction is performed based on the body tracking data of several samples of each behavioral phenotype sample trajectory to obtain three-dimensional behavior representation data. Behavioral clustering is performed on each of the three-dimensional behavioral representation data to obtain multiple behavioral data clusters; Each of the behavioral data clusters is labeled to obtain the behavioral phenotype labeled data corresponding to the trajectory of each behavioral phenotype sample.

[0009] According to some embodiments of this application, the step of supervising the training of the classification detection model based on each of the behavioral phenotype sample trajectories and the behavioral phenotype annotation data corresponding to each of the behavioral phenotype sample trajectories to obtain the pre-trained phenotype classification model includes: The behavioral phenotype sample trajectory is input into the classification detection model for classification training to obtain the classification training results of this iteration. The behavioral phenotype annotation data corresponding to the behavioral phenotype sample trajectory is compared with the classification training results to obtain the training bias data for this round. The model parameters of the classification and detection model are adjusted based on the training bias data of this round to update the classification and detection model; Return to the previous step and input the behavioral phenotype sample trajectory into the updated classification and detection model to obtain the classification training result for the next iteration, until the classification and detection model meets the preset training expectation conditions, and obtain the pre-trained phenotype classification model.

[0010] According to some embodiments of this application, the step of detecting social disorder behaviors by using a pre-trained phenotypic classification model to analyze the behavioral trajectories of the experimental animals and control animals, and obtaining animal social disorder behavior detection data, includes detecting social disorder behaviors based on the behavioral trajectories of the experimental animals, specifically including: Using the phenotypic classification model, cluster analysis is performed on the experimental animal's behavioral trajectory in multiple behavioral data clusters to determine the target behavioral cluster. Based on the behavioral phenotypic annotation data corresponding to the target behavior cluster, the animal social disorder behavior detection data is determined.

[0011] According to some embodiments of this application, the step of detecting social disorder behavior by using a pre-trained phenotypic classification model to analyze the behavioral trajectories of the experimental animals and control animals, and obtaining animal social disorder behavior detection data, includes: The experimental animal's behavioral trajectory is input into the phenotypic classification model for phenotypic identification to obtain the experimental animal's experimental behavioral phenotype. The behavioral trajectories of the control animals are input into the phenotypic classification model for phenotypic identification, thereby obtaining the control behavioral phenotypes of the control animals. The difference between the behavioral trajectories of the experimental animals and the control animals is calculated using the phenotypic classification model to obtain the experimental quantitative difference index between the experimental animals and the control animals. Based on the experimental behavioral phenotype, the control behavioral phenotype, and the experimental quantitative difference index, animal social disorder behavior detection data are generated.

[0012] According to some embodiments of this application, the step of extracting the experimental animal's behavioral trajectory from the multi-view activity footage includes: Based on the experimental multi-view activity footage corresponding to each shooting angle, two-dimensional body trajectory tracking is performed on the experimental animal to obtain a two-dimensional behavioral trajectory. Obtain the camera calibration parameters corresponding to each of the aforementioned shooting angles; The camera calibration parameters and the two-dimensional behavior trajectory are matched one-to-one by multiple sets of three-dimensional modeling processing to obtain the three-dimensional behavior trajectory. The three-dimensional behavioral trajectory is identified by a pre-trained animal posture estimation model to obtain the behavioral trajectory of the experimental animal.

[0013] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the animal social disorder behavior detection method as described in any one of the embodiments of the first aspect of this application.

[0014] Fourthly, embodiments of this application provide a computer-readable storage medium storing a program that is executed by a processor to implement the animal social disorder behavior detection method as described in any one of the embodiments of the first aspect of this application.

[0015] The animal social disorder behavior detection device, method, electronic device, and medium according to the embodiments of this application have at least the following beneficial effects: According to an embodiment of this application, the animal social disorder behavior detection device includes an observation box, an image acquisition device, and a data processing unit. The inner surface of the observation box forms an observation cavity, providing space for the animal's behavioral activities. The observation box is divided into a first sub-chamber and a second sub-chamber, which are isolated from each other. The first sub-chamber is used to house experimental animals, and the second sub-chamber is used to simultaneously house control animals under the same environmental conditions as the first sub-chamber. The experimental animals and control animals are the same species exhibiting different behavioral behaviors. The image acquisition device includes multiple cameras positioned from different shooting angles, each camera corresponding to a shooting angle. The first multi-view image acquisition device in the first sub-chamber is used to acquire experimental multi-view activity images of the experimental animals, and the second multi-view image acquisition device in the second sub-chamber is used to acquire control multi-view activity images of the control animals. The data processing unit is connected to the first and second multi-view image acquisition devices. It extracts the behavioral trajectories of experimental animals from the multi-view experimental activity footage and the behavioral trajectories of control animals from the control multi-view activity footage. Then, a pre-trained phenotypic classification model is used to detect social disorder behaviors in both the experimental and control animal behavioral trajectories, resulting in data on animal social disorder behaviors. This allows for efficient detection of animal behavior, enabling better quantitative analysis of animal behavior.

[0016] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0017] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a schematic diagram of the structure of the animal social disorder behavior detection device provided in the embodiments of this application; Figure 2 A schematic flowchart of an animal social disorder behavior detection method provided in this application embodiment; Figure 3 Another schematic diagram of the animal social disorder behavior detection method provided in the embodiments of this application; Figure 4 Another schematic diagram of the animal social disorder behavior detection method provided in the embodiments of this application; Figure 5 Another schematic diagram of the animal social disorder behavior detection method provided in the embodiments of this application; Figure 6 Another schematic diagram of the animal social disorder behavior detection method provided in the embodiments of this application; Figure 7Another schematic diagram of the animal social disorder behavior detection method provided in the embodiments of this application; Figure 8 Another schematic diagram of the animal social disorder behavior detection method provided in the embodiments of this application; Figure 9 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0018] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0019] In the description of this application, "several" means one or more, "more than" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. The use of "first" and "second" in the description is merely for distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.

[0020] In the description of this application, it should be understood that the orientation descriptions, such as up, down, left, right, front, and back, are based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0021] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0022] In the description of this application, it should be noted that, unless otherwise explicitly defined, terms such as "setting," "installation," and "connection" should be interpreted broadly. Those skilled in the art can reasonably determine the specific meaning of the above terms in this application based on the specific content of the technical solution. Furthermore, the identification of specific steps in the following text does not imply a limitation on the order of steps or execution logic. The execution order and logic between each step should be understood and inferred from the content described in the embodiments.

[0023] Autism spectrum disorders, as a neurodevelopmental disorder with a continuously rising incidence, rely heavily on accurate animal model behavioral assessments for understanding their pathological mechanisms and developing intervention strategies. Traditional behavioral assessment methods mainly depend on manual observation or two-dimensional video analysis. Typical paradigms include open-field tests, three-box social interaction tests, social interaction tests, and object recognition tests. These methods semi-quantitatively evaluate mice's social approach and avoidance, ability to explore novel environments, and stereotyped repetitive behaviors by recording the animal's positional trajectory in a specific scenario and the frequency and duration of contact with target objects or peers.

[0024] However, such methods have limitations in terms of data acquisition dimensions and behavioral decoding depth. They can only achieve planar trajectory tracking from a single camera perspective, obtaining coarse-grained spatial information about "where the animal is active," but cannot accurately identify the fine semantics of "what behavior is being performed." Especially in the quantification of stereotyped behaviors, due to the lack of continuous capture of three-dimensional movements of key body parts, it is difficult to objectively determine the start time, duration, and evolutionary pattern of abnormal repetitive movements. Furthermore, it is impossible to distinguish specific behavioral subtypes such as grooming, jumping, and rotating, resulting in evaluation results mixed with a large amount of subjective judgment bias. The consistency coefficient between different experimenters is usually lower than the benchmark requirements for reproducibility studies.

[0025] More importantly, the evaluation dimensions of existing technical architectures are highly fragmented. Social behavior assessment only focuses on the statistics of simple contact events between animals, ignoring rich social microstructures such as gesture communication, synchronized movement, and olfactory detection; stereotyped behavior detection is limited to a few isolated indicators such as the number of times animals are rolled over and the duration of grooming. This one-sidedness makes it difficult for the evaluation data to fully reflect the core phenotypic characteristics of autism model mice, reducing the effectiveness of detecting social impairment behaviors in animals.

[0026] Therefore, how to efficiently detect animal behavior and achieve quantitative analysis of animal behavior has become a major problem that urgently needs to be solved in the industry.

[0027] This application aims to address at least one of the technical problems existing in the prior art. To this end, this application proposes an animal social behavior detection device and method, electronic device, and medium, which can efficiently detect animal behavior activities to better achieve quantitative analysis of animal behavior activities.

[0028] The following explanation is based on the accompanying drawings.

[0029] The animal social disorder behavior detection device according to the embodiments of this application may include: The observation box has an inner side that forms an observation cavity, which provides space for the animals to move around. The observation box is divided into two isolated sub-chambers: the first sub-chamber is used to house the experimental animals, and the second sub-chamber is used to house the control animals under the same environmental conditions as the first sub-chamber. The experimental animals and the control animals are the same species that exhibit different behaviors. The image acquisition device includes multiple cameras set from different shooting angles, used to capture the behavior and activities of animals from multiple shooting angles, with each camera corresponding to a shooting angle. The image acquisition device set in the first sub-chamber is a first multi-view image acquisition device, which is used to acquire experimental multi-view activity images of experimental animals. The image acquisition device set in the second sub-chamber is a second multi-view image acquisition device, which is used to acquire control multi-view activity images of the control animals. The data processing unit is connected to the first multi-view image acquisition device and the second multi-view image acquisition device. The data processing unit is used to extract the behavioral trajectory of the experimental animal from the experimental multi-view activity picture and extract the behavioral trajectory of the control animal from the control multi-view activity picture. Then, the social impairment behavior detection is performed on the experimental animal behavioral trajectory and the control animal behavioral trajectory through a pre-trained phenotypic classification model to obtain animal social impairment behavior detection data.

[0030] Reference Figure 1 The image shown is a schematic diagram of an animal social disorder behavior detection device. Figure 1 As can be seen, the entire apparatus is surrounded by an observation box, which is internally divided into two sub-chambers: a first sub-chamber and a second sub-chamber. These two sub-chambers are isolated from each other, and each sub-chamber is used to house animals. The first sub-chamber is used to house experimental animals, while the second sub-chamber is used to house control animals. This design allows the experimenter to simultaneously observe and record the behavior of both experimental and control animals under controlled environmental conditions.

[0031] Within each sub-chamber, a multi-view image acquisition device is installed. This device consists of multiple cameras used to capture the animal's behavior from different angles. The multi-view image acquisition device in the first sub-chamber is called the first multi-view image acquisition device, and it is responsible for acquiring multi-view images of the experimental animals. Correspondingly, the multi-view image acquisition device in the second sub-chamber is called the second multi-view image acquisition device, and it is responsible for acquiring multi-view images of the control animals. Each camera corresponds to a specific shooting angle, ensuring comprehensive capture of the animal's three-dimensional behavior.

[0032] The video data captured by these cameras is transmitted to the data processing unit. This unit, a crucial component of the device, receives video streams from the two multi-view image acquisition devices and extracts the animals' behavioral trajectories. Specifically, the data processing unit analyzes the multi-view footage of the experimental animals to extract their behavioral trajectories; simultaneously, it analyzes the multi-view footage of control animals to extract their behavioral trajectories. This behavioral trajectory data is then used to detect social disorder behaviors.

[0033] The entire device is designed to efficiently detect and analyze social behavior disorders in animals through automation. By acquiring multi-view images, the device obtains a comprehensive view of the animal's behavior, and the data processing unit extracts key behavioral information from these views. This method not only improves the efficiency of data acquisition but also reduces the subjectivity of human observation, thereby enhancing the accuracy and reliability of behavioral analysis. In this way, researchers can better understand and assess social behavior disorders in animals, providing crucial data support for related research.

[0034] It is important to note that the core problem with traditional methods for detecting social disorder behaviors in animals lies in the limited dimensions of data collection and insufficient behavioral decoding capabilities. These methods generally rely on single-camera perspectives for planar trajectory tracking, only acquiring information about the animal's position and movement in two-dimensional space. They cannot capture the three-dimensional movement details of various joints, resulting in behavioral analysis remaining at a coarse-grained level of "where the animal is moving," making it difficult to deeply identify the semantic information of "what fine movements the animal is performing." Especially in the quantification of stereotyped behaviors, the lack of continuous three-dimensional motion capture for specific behavioral subtypes such as grooming, jumping, and rotating makes it impossible to accurately determine the onset time, duration, and dynamic evolution pattern of abnormal repetitive movements. Furthermore, manual observation or semi-automatic video analysis introduces a large amount of subjective judgment bias, and the consistency of ratings among different experimenters is difficult to meet the benchmark requirements for reproducibility studies. This makes the detection data unable to comprehensively and objectively reflect the core phenotypic characteristics of autism model mice, thus limiting the accuracy and scientific validity of behavioral detection.

[0035] The "difference" between experimental animals and control animals refers to fundamental differences in their genetic background or experimental intervention conditions. Specifically, experimental animals may be gene knockout autism model mice or subjects treated with drugs or environmental factors, while control animals are wild-type healthy mice of the same genetic lineage or baseline samples that have not received special treatment. This difference is a prerequisite for conducting disease mechanism research or efficacy evaluation. Without this pre-established biological difference, subsequent detection of social impairment behaviors would lose its scientific significance. Therefore, the difference in physiological or pathological states between the two is precisely the target variable that this device needs to detect and quantify.

[0036] To achieve this design goal, the observation chamber employs a dual-chamber parallel layout in its physical structure. While the first and second sub-chambers are separated by partitions to prevent interference between animals, they share the same macroscopic environmental system. The internal temperature of both chambers is regulated by a unified environmental control module. The lighting system uses synchronized LED light source arrays to ensure uniform illuminance distribution. The chamber materials are consistent with the background color to eliminate visual environmental differences. Even airflow circulation and olfactory stimulation are balanced through a symmetrical ventilation system design. This design allows experimental and control animals to be simultaneously recorded in identical temperature, humidity, light cycle, background color, and acoustic environments, ensuring that any observed differences in behavioral trajectories can only be attributed to the animal's own phenotypic characteristics, rather than fluctuations in the external environment.

[0037] The technical advantage of this synchronous acquisition mechanism lies in the fact that the experimental multi-view activity images and control multi-view activity images received by the data processing unit are not only perfectly aligned in terms of timestamps, but more importantly, they maintain a high degree of consistency in parameters at the data acquisition level, such as environmental background, lighting conditions, and image quality. This provides high-quality comparable data pairs for subsequent phenotypic classification models, enabling the models to learn based on purely biological differences without the need for additional learning on how to eliminate environmental noise. In other words, through rigorous isolation and environmental uniformity at the hardware level, the device solidifies the internal validity assurance mechanism of the experimental design into a physical structure, thereby significantly improving the scientific credibility and cross-laboratory reproducibility of the social disorder behavior detection results.

[0038] The animal social disorder behavior detection device of this application fundamentally solves the problem of environmental variable control through a dual-chamber parallel design of the observation chamber. The observation chamber is divided into two isolated sub-chambers, physically independent but sharing identical environmental conditions such as temperature, humidity, light, and background noise. The first sub-chamber houses the experimental animals, while the second sub-chamber simultaneously houses the control animals, allowing for behavioral observation in a completely synchronized time dimension. This design strictly separates the biological differences between experimental and control animals from environmental interference factors, ensuring that any observed behavioral differences can only be attributed to the animal's own phenotypic characteristics, rather than fluctuations in the external environment. Compared to the environmental bias risks associated with separate time-sharing and location-based testing of experimental and control groups in traditional methods, the dual-chamber synchronous acquisition mechanism provides a highly comparable data foundation for subsequent behavioral trajectory extraction, significantly improving the internal validity and cross-experimental reproducibility of the detection results.

[0039] At the data acquisition level, this device employs a multi-view image acquisition system, overcoming the planar limitations of a single camera perspective. Each sub-chamber is equipped with a first and a second multi-view image acquisition system, each containing multiple cameras positioned from different shooting angles, each camera corresponding to an independent shooting angle. This multi-camera array layout can simultaneously capture animal behavior from multiple directions, constructing an observation network covering three-dimensional space. When experimental animals exhibit complex movements such as circling, grooming, or standing on their hind limbs, image data from different perspectives can be fused into a three-dimensional spatial trajectory through subsequent processing, effectively eliminating self-occlusion and depth ambiguity issues. Compared to traditional two-dimensional video analysis, which can only acquire planar projection information, multi-view activity footage provides complete three-dimensional kinematic data for behavioral trajectory extraction, making continuous tracking of key points on the animal's body (such as the head, neck, limbs, and tail) possible. This allows for accurate identification of fine motor subtypes such as grooming, jumping, and rotating, as well as their spatiotemporal evolution characteristics, achieving a technological leap from coarse-grained position tracking to fine-grained semantic decoding of movements.

[0040] Regarding the objectivity and automation of behavior detection, this device completely replaces subjective human scoring through the collaborative work of a data processing unit and a pre-trained phenotypic classification model. The data processing unit is directly connected to the first and second multi-view image acquisition devices, performing a unified computational processing flow on both experimental and control multi-view activity footage. This process sequentially completes steps such as animal body keypoint detection, 3D trajectory reconstruction, and kinematic parameter extraction, generating standardized behavioral trajectory data. The pre-trained phenotypic classification model undergoes supervised learning based on a large number of labeled socially awkward behavior samples and normal behavior samples, possessing the ability to automatically identify phenotypes such as social avoidance anomalies, stereotyped repetitive actions, and atypical exploration patterns. By comparing the distribution differences of experimental and control animal behavioral trajectories in the feature space, the model outputs quantitative socially awkward behavior detection data, avoiding cognitive biases and scoring inconsistencies inherent in human observation. This end-to-end automated analysis chain not only significantly improves detection efficiency but also ensures consistency of evaluation standards across different batches of experiments through standardized processing of the algorithm model, establishing the quantitative analysis of animal socially awkward behavior on an objective and repeatable algorithmic foundation.

[0041] In summary, the embodiments of this application form a hardware and software collaborative technical solution through the synchronous control of the dual-cavity environment of the observation box, the three-dimensional data capture capability of the multi-view image acquisition device, and the automated intelligent analysis of the data processing unit and the phenotypic classification model. This solution systematically solves multiple limitations of traditional methods in terms of environmental variable control, data acquisition dimensions, behavioral decoding depth, and subjective bias. It achieves a technological upgrade from planar trajectory recording to three-dimensional action recognition, and from manual qualitative scoring to algorithmic quantitative analysis, providing an efficient, accurate, and objective automated technical platform for detecting animal social behavior disorders.

[0042] Reference Figure 2 The detection method for animal social disorder behaviors according to the embodiments of this application, applied to the animal social disorder behavior detection device of the embodiments of this application, may include: Step S201: The experimental animal housed in the first sub-chamber is captured by the first multi-view image acquisition device to obtain multi-view images of the experimental animal’s experimental activities. Step S202: The control animal placed in the second sub-chamber is captured by the second multi-view image acquisition device to obtain the control animal's multi-view activity image. Step S203: Extract the behavioral trajectories of experimental animals from the multi-view activity footage of the experiment; Step S204: Extract the behavioral trajectory of the control animal from the multi-view activity footage; Step S205: Social disorder behavior detection is performed on the behavioral trajectories of experimental animals and control animals using a pre-trained phenotypic classification model to obtain animal social disorder behavior detection data.

[0043] The animal social disorder behavior detection method proposed in this application is a highly automated and precise detection process designed to assess animal behavioral characteristics through advanced image acquisition technology and data processing algorithms. The first step of this method, step S201, involves using a first multi-view image acquisition device to capture the behavior of an experimental animal placed in a first sub-chamber. This device consists of multiple cameras that capture images of the experimental animal from different angles, thereby obtaining a set of multi-view activity images. These images provide information about the experimental animal's activity in three-dimensional space, including its position, posture, and movements.

[0044] Next, in step S202, a second multi-view image acquisition device is used to acquire similar multi-view images of the control animal placed in the second sub-chamber. This device has the same structure and function as the first multi-view image acquisition device, ensuring that the experimental and control animals are recorded under the same conditions. In this way, researchers can obtain multi-view images of the control animals' activities, which also contain rich three-dimensional spatial behavioral information. This design allows researchers to compare behavioral differences between experimental and control animals while controlling experimental conditions.

[0045] In steps S203 and S204, the data processing unit begins to function. It first extracts the behavioral trajectories of the experimental animals from the multi-view video footage, and then extracts the behavioral trajectories of the control animals from the control multi-view video footage. This step involves complex image processing and computer vision techniques, such as keypoint detection, motion tracking, and 3D reconstruction. These techniques allow for the extraction of the animal's motion parameters, including position, velocity, and acceleration, from the raw video data, thereby obtaining a precise behavioral trajectory.

[0046] Finally, in step S205, the researchers use a pre-trained phenotypic classification model to analyze these behavioral trajectories. This model, developed based on machine learning or deep learning algorithms, is capable of identifying and distinguishing different behavioral patterns. By analyzing the behavioral trajectories of experimental and control animals, the model can detect whether the animals exhibit social impairment behaviors and generate corresponding detection data. This data includes not only the judgment results of whether the animals exhibit social impairment, but may also include detailed information such as the degree of behavioral abnormality and the frequency of specific behaviors.

[0047] In summary, the animal social behavior detection method of this application, by combining multi-view image acquisition technology and advanced data processing algorithms, provides an efficient, accurate, and automated means of animal behavior analysis. This method can significantly improve researchers' understanding and assessment of animal social behavior disorders, providing a valuable tool for related research fields.

[0048] Reference Figure 3 According to some embodiments of this application, before detecting social disorder behavior by using a pre-trained phenotypic classification model to analyze the behavioral trajectories of experimental animals and control animals, and obtaining animal social disorder behavior detection data, the method further includes pre-training the phenotypic classification model, which may specifically include: Step S301: Determine the original classification and detection model; Step S302: Obtain the behavior detection training dataset; wherein, the behavior detection training dataset includes multiple behavior phenotype sample trajectories and behavior phenotype annotation data corresponding to each behavior phenotype sample trajectory; Step S303: Based on the trajectory of each behavioral phenotype sample and the behavioral phenotype annotation data corresponding to each behavioral phenotype sample trajectory, supervised training is performed on the classification detection model to obtain a pre-trained phenotype classification model.

[0049] In some embodiments of this application, to ensure the accuracy and reliability of detecting social disorder behaviors in animals, a pre-training process is required before using the pre-trained phenotypic classification model for detection. This process is crucial because it directly affects the accuracy of the model's detection results.

[0050] First, in step S301, the original classification and detection model needs to be determined. This model is the foundation of behavior detection; it could be a deep learning model, such as a convolutional neural network (CNN) or a recurrent neural network (RNN), or a traditional machine learning model, such as a support vector machine (SVM) or a random forest. The choice of model depends on several factors, including the expected detection accuracy, the availability of computational resources, and the characteristics of the dataset. After determining the model, its architecture needs to be designed, including the number of layers, the number of neurons in each layer, and the choice of activation functions, so that the model can effectively learn features from the input data.

[0051] Next, in step S302, it is necessary to obtain the behavior detection training dataset. This dataset is crucial for training the model. It includes multiple behavioral phenotypic sample trajectories and corresponding behavioral phenotypic annotation data for each trajectory. Behavioral phenotypic sample trajectories refer to the movement trajectories of experimental or control animals under specific behaviors. These trajectories can be two-dimensional or three-dimensional, depending on the configuration of the image acquisition device. The behavioral phenotypic annotation data is the interpretation of these trajectories, including the type of behavior the animal is performing, such as social interaction, exploration, repetitive actions, etc. This annotation data is usually provided manually by experts to ensure the accuracy and reliability of the data.

[0052] Finally, in step S303, the classification and detection model is trained under supervised supervision based on the collected behavioral phenotypic sample trajectories and corresponding behavioral phenotypic annotation data. Supervised training is a machine learning method in which the model learns how to correctly classify new data by comparing its predictions with the actual annotation data. During training, the model continuously adjusts its internal parameters to minimize the difference between the predictions and the actual annotations. This process may include multiple iterations, each of which makes the model's predictions more accurate. After sufficient training, the model will be able to identify and classify new behavioral trajectory data, thus obtaining a pre-trained phenotypic classification model.

[0053] This pre-trained phenotypic classification model can then be used to detect social disorder behaviors in animals. When behavioral trajectory data from new experimental and control animals are collected and input into the model, the model can classify these trajectories based on the features and patterns learned during training, determine whether the animals exhibit social disorder behaviors, and provide corresponding detection data. Such automated detection not only improves detection efficiency but also reduces interference from human factors, making research results more objective and reliable.

[0054] Reference Figure 4 According to some embodiments of this application, obtaining a behavior detection training dataset may include: Step S401: Obtain multiple behavioral phenotype sample trajectories; wherein each behavioral phenotype sample trajectory corresponds to a sample animal's behavioral phenotype. Step S402: Perform two-dimensional body tracking on the trajectory of each behavioral phenotypic sample to obtain several sample body tracking data. Step S403: Based on the body tracking data of several samples of each behavioral phenotype sample trajectory, perform three-dimensional behavior reconstruction to obtain three-dimensional behavior representation data. Step S404: Perform behavioral clustering on each three-dimensional behavioral representation data to obtain multiple behavioral data clusters; Step S405: Label each row of data clusters to obtain the behavior phenotype labeling data corresponding to the trajectory of each row phenotype sample.

[0055] In some embodiments of this application, obtaining a behavioral detection training dataset is a crucial task. This involves extracting and processing data from the behavioral trajectories of experimental animals for training a phenotypic classification model. The first step in this process, step S401, is to obtain multiple behavioral phenotypic sample trajectories. These sample trajectories reflect the behavioral performance of the sample animals under specific experimental conditions, with each sample trajectory corresponding to a specific behavioral phenotype of a sample animal. These behavioral phenotypes may include social interaction, exploratory behavior, repetitive actions, etc., and they form the basis for studying animal social disorder behaviors.

[0056] After acquiring the behavioral phenotypic sample trajectories, step S402 requires two-dimensional body tracking for each sample trajectory. The purpose of this step is to capture the animal's motion details in a two-dimensional plane, including the position and movement path of various body parts. By using computer vision techniques, such as background subtraction, edge detection, or deep learning algorithms, the animal's contours or key points can be extracted from the video data, and the changes of these points over time can be tracked. The resulting sample body tracking data provides the foundation for subsequent three-dimensional reconstruction.

[0057] It should be noted that step S403 is the process of reconstructing three-dimensional behavior based on two-dimensional tracking data. Since two-dimensional tracking only provides planar information, while animal behavior is an activity in three-dimensional space, it is necessary to convert the two-dimensional data into three-dimensional behavioral representation data. This step typically involves using multi-view images and geometric algorithms to reconstruct the animal's three-dimensional posture and trajectory. By integrating two-dimensional tracking data from different cameras, the positions of various parts of the animal's body in three-dimensional space can be calculated, thus obtaining a complete three-dimensional behavioral representation.

[0058] In step S404, behavioral clustering is performed on the three-dimensional behavioral representation data to identify and classify different behavioral patterns. Behavioral clustering is an unsupervised learning method that groups similar behavioral samples into a single class based on the inherent structure of the data, forming multiple behavioral data clusters. These clusters represent different behavioral types, such as exploration, rest, and social interaction. Cluster analysis helps extract meaningful patterns from complex behavioral data and provides structured data for further behavioral analysis and model training.

[0059] Finally, in step S405, each behavioral data cluster needs to be labeled to generate behavioral phenotypic labeled data. This step is typically performed by domain experts who assign a behavioral label, such as "social interaction" or "repetitive action," to each cluster based on the clustering results and behavioral knowledge. This labeled data corresponds to the original behavioral phenotypic sample trajectories, forming a labeled dataset used to train the phenotypic classification model. This ensures that the model learns the correct behavioral classification criteria during training, thereby improving the model's accuracy and reliability in practical applications.

[0060] Reference Figure 5 According to some embodiments of this application, a pre-trained phenotypic classification model is obtained by supervising the training of a classification detection model based on the trajectory of each behavioral phenotypic sample and the behavioral phenotypic annotation data corresponding to each behavioral phenotypic sample trajectory, which may include: Step S501: Input the behavioral phenotype sample trajectory into the classification detection model for classification training to obtain the classification training results of this iteration. Step S502: Compare the behavioral phenotype annotation data corresponding to the behavioral phenotype sample trajectory with the classification training results to obtain the training bias data for this round. Step S503: Adjust the model parameters of the classification and detection model based on the training bias data of this round to update the classification and detection model; Step S504: Return to the previous step and input the behavioral phenotype sample trajectory into the updated classification and detection model to obtain the classification training results for the next iteration, until the classification and detection model meets the preset training expectation conditions and obtains the pre-trained phenotype classification model.

[0061] In some embodiments of this application, the supervised training process for the classification detection model is an iterative optimization process aimed at improving the model's classification accuracy for behavioral phenotypic sample trajectories by continuously adjusting the model parameters. This process begins in step S501, where the behavioral phenotypic sample trajectories are input into the classification detection model for training. In this step, the model analyzes the input behavioral samples based on its current parameter settings and attempts to classify them into predefined behavioral categories. After completing this round of classification, the model produces a classification training result, which is the model's predicted output based on the input data.

[0062] Next, in step S502, the behavioral phenotypic labeled data corresponding to the behavioral phenotypic sample trajectories needs to be compared with the classification training results generated by the model. The behavioral phenotypic labeled data are accurate classifications of the sample trajectories by experts based on behavioral knowledge; they represent the true behavioral category of each sample. By comparing the model's prediction results with the actual labeled data, the bias data for this round of training can be calculated. This bias data reflects the difference between the model's predictions and the actual labels, and is a key indicator for evaluating model performance.

[0063] In step S503, based on the calculated training bias data of this round, the model parameters of the classification and detection model need to be adjusted. This step is achieved through optimization algorithms, such as gradient descent or other machine learning optimization algorithms. The optimization algorithm guides the direction and magnitude of model parameter adjustment based on the bias data, aiming to reduce the difference between model predictions and true labels. In this way, the model gradually learns more accurate classification rules in each iteration.

[0064] Finally, in step S504, the behavioral phenotypic sample trajectories need to be input again into the updated classification and detection model to obtain the classification training results for the next iteration. This process is repeated, with each iteration using the updated model parameters for training, until the model's performance meets the preset training expectations. These expectations may include achieving a certain accuracy, recall, or other performance metrics. Once the model meets these conditions, it can be considered that the model has been sufficiently trained, and the model at this point is the pre-trained phenotypic classification model, which can be used for actual animal social behavior detection tasks.

[0065] In summary, this supervised training process is one of continuous trial and error and optimization. By constantly comparing the model's predictions with real labeled data, the model parameters are gradually adjusted, ultimately resulting in a pre-trained phenotypic classification model capable of accurately classifying behavioral phenotypic sample trajectories. This model can serve as a core component of animal social behavior detection devices, providing researchers with automated behavioral analysis tools.

[0066] Reference Figure 6 According to some embodiments of this application, social disorder behavior detection is performed on the behavioral trajectories of experimental animals and control animals using a pre-trained phenotypic classification model to obtain animal social disorder behavior detection data. This includes detecting social disorder behavior based on the behavioral trajectories of experimental animals, specifically including: Step S601: Using a phenotypic classification model, cluster analysis is performed on the experimental animal behavioral trajectories in multiple behavioral data clusters to determine the target behavioral cluster. Step S602: Determine the animal social disorder behavior detection data based on the behavioral phenotypic annotation data corresponding to the target behavior cluster.

[0067] In some embodiments of this application, a key step is to use a pre-trained phenotypic classification model to detect social impairment behaviors in the behavioral trajectories of experimental animals, aiming to identify and quantify such behaviors. This process begins in step S601, where the phenotypic classification model is used to perform cluster analysis on the behavioral trajectories of the experimental animals. Cluster analysis is an unsupervised learning method that groups similar behaviors into a single category based on the characteristics of the behavioral trajectories, forming multiple clusters of behavioral data. These clusters represent different behavioral patterns, such as social interaction, exploration, and repetitive actions. In this way, the model can identify the behavioral category to which the experimental animal's behavioral trajectory most likely belongs.

[0068] During cluster analysis, the phenotypic classification model calculates the similarity or distance between the experimental animal's behavioral trajectories and each behavioral data cluster. These similarities or distances can be calculated based on various features, such as the shape, speed, frequency, and duration of the behavioral trajectory. Based on these calculations, the model assigns the experimental animal's behavioral trajectories to the most similar behavioral data clusters, thus determining the target behavioral cluster. This step is automated, reducing interference from subjective human judgment and improving the objectivity and consistency of behavioral classification.

[0069] Next, in step S602, the detection data for animal social disorder behaviors is determined based on the behavioral phenotypic annotation data corresponding to the target behavior clusters. During the model training phase, each behavior data cluster is assigned a behavioral phenotypic annotation, which is provided by domain experts and represents the behavior category of each cluster. When the experimental animal's behavioral trajectory is assigned to a target behavior cluster, the model refers to the behavioral phenotypic annotation data corresponding to that cluster to determine the animal's behavior category. This step involves matching the clustering results with known behavior categories to obtain the detection data for animal social disorder behaviors.

[0070] These test data may include assessments of whether the animal exhibits socially dysfunctional behaviors, as well as detailed information such as the severity of the behavior and the frequency of specific behaviors. This data is extremely valuable to researchers because it provides a quantitative and objective method for evaluating socially dysfunctional behaviors in animals. In this way, researchers can more accurately understand the behavioral characteristics of animals, providing a scientific basis for further research and treatment.

[0071] In summary, the process of detecting social disorder behaviors in laboratory animals by using pre-trained phenotypic classification models is a highly automated and precise process. It utilizes advanced machine learning and data analysis techniques to transform complex behavioral data into meaningful behavioral categories and detection results, providing a new tool for animal behavior research and the assessment of social disorder behaviors.

[0072] Reference Figure 7 According to some embodiments of this application, social disorder behavior detection data is obtained by using a pre-trained phenotypic classification model to analyze the behavioral trajectories of experimental animals and control animals, which may include: Step S701: Input the experimental animal behavior trajectory into the phenotypic classification model for phenotypic identification to obtain the experimental animal's experimental behavior phenotype; Step S702: Input the behavioral trajectory of the control animal into the phenotypic classification model for phenotypic identification to obtain the control behavioral phenotype of the control animal; Step S703: The difference between the behavioral trajectories of experimental animals and control animals is calculated using a phenotypic classification model to obtain the experimental quantitative difference index between experimental animals and control animals. Step S704: Generate animal social disorder behavior detection data based on experimental behavioral phenotypes, control behavioral phenotypes, and experimental quantitative difference indicators.

[0073] In some embodiments of this application, the process of detecting social impairment behaviors in the behavioral trajectories of experimental animals and control animals using a pre-trained phenotypic classification model is a systematic and automated analysis process. The first step of this process, step S701, involves inputting the behavioral trajectories of the experimental animals into the phenotypic classification model for phenotypic identification. During this process, the model analyzes the behavioral trajectories of the experimental animals based on the features and patterns learned during the training phase and derives an experimental behavioral phenotype. This phenotype reflects the behavioral characteristics of the experimental animals, such as social interaction and exploratory behavior, and can reveal whether they exhibit social impairment behaviors.

[0074] Next, in step S702, the behavioral trajectories of the control animals are input into the same phenotypic classification model for phenotypic identification to obtain the control behavioral phenotype. Control animals are typically those with similar genetic backgrounds to the experimental animals but raised under normal conditions; their behavioral trajectories are used as a reference to facilitate comparison and evaluation of behavioral changes in the experimental animals. The control behavioral phenotype provides a benchmark to help researchers understand whether the behavior of the experimental animals deviates from the normal range.

[0075] In step S703, a phenotypic classification model is used to calculate the differences between the behavioral trajectories of experimental animals and control animals, obtaining quantitative difference indicators between the experimental animals and control animals. This step involves comparing experimental and control behavioral phenotypes and calculating the differences between them. These difference indicators can be quantitative comparisons of behavioral characteristics, such as the frequency of social interaction and the duration of exploratory behavior, or they can be statistical quantities output by the model, such as classification confidence and probability distribution differences. These quantitative difference indicators provide specific numerical basis for assessing social impairment behaviors in experimental animals.

[0076] Finally, in step S704, animal social disorder behavior detection data are generated based on experimental behavioral phenotypes, control behavioral phenotypes, and experimental quantitative difference indicators. This detection data integrates the behavioral characteristics of experimental and control animals and highlights the differences between them. This data may include detailed information such as the judgment of whether the experimental animal exhibits social disorder behavior, the degree of behavioral abnormality, and the frequency of specific behaviors. This information is extremely valuable to researchers because it provides a quantitative and objective method to assess animal social disorder behavior, contributing to a deeper understanding of animal behavioral characteristics and the biological mechanisms of social disorders.

[0077] In summary, this process of detecting social disorder behaviors in experimental and control animals by using a pre-trained phenotypic classification model is a highly automated and precise analytical procedure. It utilizes advanced machine learning and data analysis techniques to transform complex behavioral data into meaningful behavioral phenotypes and quantitative indicators of difference, providing a new tool for animal behavior research and the assessment of social disorder behaviors.

[0078] Reference Figure 8 According to some embodiments of this application, extracting the behavioral trajectory of experimental animals from multi-view experimental video footage may include: Step S801: Based on the experimental multi-view activity images corresponding to each shooting angle, perform two-dimensional body trajectory tracking on the experimental animal to obtain a two-dimensional behavioral trajectory. Step S802: Obtain the camera calibration parameters corresponding to each shooting angle; Step S803: Perform three-dimensional modeling processing on multiple sets of one-to-one corresponding camera calibration parameters and two-dimensional behavior trajectories to obtain three-dimensional behavior trajectories; Step S804: The three-dimensional behavioral trajectory is identified by a pre-trained animal posture estimation model to obtain the experimental animal's behavioral trajectory.

[0079] In some embodiments of this application, extracting the behavioral trajectory of the experimental animal from multi-view activity footage is a crucial step, involving the analysis and reconstruction of the animal's three-dimensional motion from videos captured from multiple perspectives. The first step of this process, step S801, involves performing two-dimensional body trajectory tracking on the experimental animal based on the multi-view activity footage corresponding to each shooting perspective, thereby obtaining a two-dimensional behavioral trajectory. In this stage, the video stream captured by each camera is analyzed independently, using computer vision techniques to identify and track the experimental animal's two-dimensional position at each perspective. This may involve using background subtraction, optical flow, or other tracking algorithms to determine the animal's position in each frame, thereby generating a series of two-dimensional coordinate points that change over time to form the two-dimensional behavioral trajectory.

[0080] Next, in step S802, it is necessary to obtain the camera calibration parameters corresponding to each shooting angle. Camera calibration is a pre-performed process used to determine the camera's intrinsic parameters (such as focal length and principal point position) and extrinsic parameters (such as the camera's position and orientation). These parameters are crucial for subsequent 3D reconstruction because they provide the geometric information needed to convert 2D image coordinates into 3D world coordinates. Typically, this involves taking a series of images at different positions and angles in front of the camera using a calibration board of known size, and then using computer vision algorithms to calculate the camera's calibration parameters from these images.

[0081] In step S803, multiple sets of one-to-one corresponding camera calibration parameters and two-dimensional behavioral trajectories are processed into three-dimensional models to obtain three-dimensional behavioral trajectories. This step is the core of the entire process. It uses two-dimensional trajectories obtained from multiple perspectives and corresponding camera calibration parameters to calculate the actual position of the experimental animal in three-dimensional space through three-dimensional geometric reconstruction algorithms (such as triangulation). This process may involve solving the parallax problem, that is, the difference between the two-dimensional projections of the same object point under different perspectives, so as to accurately reconstruct the animal's three-dimensional movement trajectory.

[0082] Finally, in step S804, the three-dimensional behavioral trajectory is identified using a pre-trained animal pose estimation model to obtain the experimental animal's behavioral trajectory. The animal pose estimation model is a deep learning model capable of identifying specific animal postures and behaviors from three-dimensional point cloud or skeleton data. This model learns from a large number of labeled animal behavior samples during the training phase, thus enabling it to identify and classify different behavioral patterns. In the application phase, the model receives the three-dimensional behavioral trajectory as input and outputs the identification results of animal behaviors, such as standing, walking, and jumping. These identification results provide a detailed description of the experimental animal's behavior and are an important foundation for detecting social disorder behaviors.

[0083] In summary, this process, by combining multi-view image acquisition, 3D reconstruction, and deep learning techniques, achieves automated extraction of precise behavioral trajectories from raw video data. This method not only improves the accuracy and efficiency of behavioral analysis but also reduces the influence of subjective human judgment, providing a new tool for animal behavior research.

[0084] Reference Figure 9 , Figure 9 This illustration shows the hardware structure of an electronic device according to another embodiment. The electronic device may include: The processor 901 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 902 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store the operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 902 and is called and executed by the processor 901 to execute the detection method for animal social disorder behavior of the embodiments of this application. The 903 input / output interface is used to implement information input and output. The communication interface 904 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 905 transmits information between various components of the device (e.g., processor 901, memory 902, input / output interface 903, and communication interface 904); The processor 901, memory 902, input / output interface 903, and communication interface 904 are connected to each other within the device via bus 905.

[0085] This application also provides a computer program product, which includes a computer program. A processor of a computer device reads and executes the computer program, causing the computer device to perform the aforementioned method for detecting social disorder behaviors in animals.

[0086] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in this disclosure and the foregoing drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “including,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatuses.

[0087] It should be understood that in this disclosure, "at least one item" means one or more, and "more than one" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0088] It should be understood that in the description of the embodiments of this application, "multiple" means two or more, "greater than", "less than", "exceeding" etc. are understood to exclude the number itself, and "above", "below", "within" etc. are understood to include the number itself.

[0089] In the several embodiments provided in this disclosure, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.

[0090] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0091] Furthermore, the functional units in the various embodiments of this disclosure can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0092] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this disclosure. The aforementioned storage medium may include: a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, and other media capable of storing program code.

[0093] It should also be understood that the various implementation methods provided in this application can be combined arbitrarily to achieve different technical effects.

[0094] The above is a detailed description of the embodiments of this disclosure. However, this disclosure is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this disclosure. All such equivalent modifications or substitutions are included within the scope defined by the claims of this disclosure.

Claims

1. A device for detecting social disorder in animals, characterized in that, include: An observation box, the inner side of which forms an observation cavity, is used to provide space for animal behavior and activity; wherein, the interior of the observation box is divided into a first sub-chamber and a second sub-chamber that are isolated from each other, the first sub-chamber is used to contain experimental animals, and the second sub-chamber is used to simultaneously contain control animals under the same environmental conditions as the first sub-chamber, wherein the experimental animals and the control animals are the same species of animals that exhibit different behavioral behaviors; An image acquisition device includes multiple cameras set from different shooting angles, used to capture the behavior and activities of animals from the multiple shooting angles, each of the cameras corresponding to one of the shooting angles; The image acquisition device installed in the first sub-chamber is a first multi-view image acquisition device, which is used to acquire experimental multi-view activity images of the experimental animal; The image acquisition device installed in the second sub-chamber is a second multi-view image acquisition device, which is used to acquire control multi-view activity images of the control animal; A data processing unit is connected to the first multi-view image acquisition device and the second multi-view image acquisition device. The data processing unit is used to extract the behavioral trajectory of the experimental animal from the experimental multi-view activity screen and extract the behavioral trajectory of the control animal from the control multi-view activity screen. Then, it uses a pre-trained phenotypic classification model to detect social disorder behavior in the experimental animal behavioral trajectory and the control animal behavioral trajectory to obtain animal social disorder behavior detection data.

2. A method for detecting social disorder behaviors in animals, characterized in that, The method applied to the animal social disorder behavior detection device of claim 1 includes: The experimental animal housed in the first sub-chamber is captured by the first multi-view image acquisition device to obtain multi-view images of the experimental animal's activities. The control animal housed in the second sub-chamber is captured by a second multi-view image acquisition device to obtain a control multi-view activity image of the control animal. Extract the behavioral trajectories of the experimental animals from the multi-view activity footage of the experiment; Extract the behavioral trajectories of the control animals from the aforementioned multi-view activity footage; Social disorder behavior detection data of the experimental animals and control animals was obtained by using a pre-trained phenotypic classification model to detect the behavioral trajectories of the experimental animals and control animals.

3. The method according to claim 2, characterized in that, Before obtaining animal social disorder behavior detection data by using a pre-trained phenotypic classification model to detect the behavioral trajectories of the experimental animals and control animals, the method further includes pre-training the phenotypic classification model, specifically including: Determine the original classification and detection model; Obtain a behavior detection training dataset; wherein, the behavior detection training dataset includes multiple behavior phenotype sample trajectories and behavior phenotype annotation data corresponding to each behavior phenotype sample trajectory; Based on the sample trajectories of each behavioral phenotype and the behavioral phenotype annotation data corresponding to each sample trajectories, the classification detection model is trained under supervision to obtain the pre-trained phenotype classification model.

4. The method according to claim 3, characterized in that, The acquisition of the behavior detection training dataset includes: Obtain multiple behavioral phenotype sample trajectories; wherein each behavioral phenotype sample trajectory corresponds to a sample behavioral phenotype of a sample animal; Two-dimensional body tracking was performed on the trajectory of each behavioral phenotype sample to obtain several sample body tracking data. Three-dimensional behavior reconstruction is performed based on the body tracking data of several samples of each behavioral phenotype sample trajectory to obtain three-dimensional behavior representation data. Behavioral clustering is performed on each of the three-dimensional behavioral representation data to obtain multiple behavioral data clusters; Each of the behavioral data clusters is labeled to obtain the behavioral phenotype labeled data corresponding to the trajectory of each behavioral phenotype sample.

5. The method according to claim 3, characterized in that, The step of supervising the training of the classification and detection model based on the behavioral phenotype sample trajectories and the behavioral phenotype annotation data corresponding to each behavioral phenotype sample trajectory to obtain the pre-trained phenotype classification model includes: The behavioral phenotype sample trajectory is input into the classification detection model for classification training to obtain the classification training results of this iteration. The behavioral phenotype annotation data corresponding to the behavioral phenotype sample trajectory is compared with the classification training results to obtain the training bias data for this round. The model parameters of the classification and detection model are adjusted based on the training bias data of this round to update the classification and detection model; Return to the previous step and input the behavioral phenotype sample trajectory into the updated classification and detection model to obtain the classification training result for the next iteration, until the classification and detection model meets the preset training expectation conditions, and obtain the pre-trained phenotype classification model.

6. The method according to claim 3, characterized in that, The step of detecting social disorder behaviors by using a pre-trained phenotypic classification model to analyze the behavioral trajectories of the experimental animals and control animals, thereby obtaining animal social disorder behavior detection data, includes detecting social disorder behaviors based on the behavioral trajectories of the experimental animals, specifically including: Using the phenotypic classification model, cluster analysis is performed on the experimental animal's behavioral trajectory in multiple behavioral data clusters to determine the target behavioral cluster. Based on the behavioral phenotypic annotation data corresponding to the target behavior cluster, the animal social disorder behavior detection data is determined.

7. The method according to claim 2, characterized in that, The social disorder behavior detection is performed on the behavioral trajectories of the experimental animals and control animals using a pre-trained phenotypic classification model, resulting in animal social disorder behavior detection data, including: The experimental animal's behavioral trajectory is input into the phenotypic classification model for phenotypic identification to obtain the experimental animal's experimental behavioral phenotype. The behavioral trajectories of the control animals are input into the phenotypic classification model for phenotypic identification, thereby obtaining the control behavioral phenotypes of the control animals. The difference between the behavioral trajectories of the experimental animals and the control animals is calculated using the phenotypic classification model to obtain the experimental quantitative difference index between the experimental animals and the control animals. Based on the experimental behavioral phenotype, the control behavioral phenotype, and the experimental quantitative difference index, animal social disorder behavior detection data are generated.

8. The method according to claim 2, characterized in that, The extraction of experimental animal behavioral trajectories from the multi-view activity footage includes: Based on the experimental multi-view activity footage corresponding to each shooting angle, two-dimensional body trajectory tracking is performed on the experimental animal to obtain a two-dimensional behavioral trajectory. Obtain the camera calibration parameters corresponding to each of the aforementioned shooting angles; The camera calibration parameters and the two-dimensional behavior trajectory are matched one-to-one by multiple sets of three-dimensional modeling processing to obtain the three-dimensional behavior trajectory. The three-dimensional behavioral trajectory is identified by a pre-trained animal posture estimation model to obtain the behavioral trajectory of the experimental animal.

9. An electronic device, characterized in that, include: The device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the detection method for animal social disorder behavior as described in any one of claims 2 to 8.

10. A computer-readable storage medium, characterized in that, The storage medium stores a program that is executed by a processor to implement the detection method for animal social disorder behaviors as described in any one of claims 2 to 8.

Images