Rat behavior monitoring system and method based on transparent cage and image processing
By combining transparent cages and high-definition cameras, combined with deep learning algorithms and contactless infrared sensors, the problem of insufficient quality of closed cages and video acquisition in the rat behavior monitoring system is solved, and the accuracy of rat behavior and physiological parameters is achieved Monitoring and evaluation improves data management efficiency.
Patent Information
- Application Number
- CN202410971534.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-19
- Publication Date
- 2024-11-08
AI Technical Summary
The existing rat behavior monitoring system has closed cages to restrict rat activities, insufficient video acquisition quality, behavior recognition algorithms cannot accurately identify complex behavior patterns, and lack physiological parameter monitoring, resulting in data errors and storage compression problems.
DeepBehaviorNet algorithm using a combination of transparent cage and high-definition cameras, integrating deep convolutional neural networks and long and short-term memory networks, combining contactless infrared sensors to monitor physiological parameters, and using multimodal fusion algorithm and attention mechanism for behavior recognition and emotional evaluation , use efficient data compression algorithm to store video data.
It improves the accuracy of behavior recognition and video data quality, realizes accurate identification of complex behavior patterns in rats and real-time monitoring of physiological parameters, provides a comprehensive health assessment, and improves data management efficiency and convenience of experimental report generation.
Smart Images

Figure CN118918529A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of animal behavior monitoring, in particular to a rat behavior monitoring system based on transparent cages and image processing and a method thereof. Background Art
[0002] In animal experiments and behavioral studies, traditional rat behavior monitoring systems usually rely on closed cages and manual observation methods. This method is not only labor-intensive, but also prone to data errors and inconsistencies due to human interference. With the development of computer technology and image processing technology, automated behavior monitoring systems based on video analysis have gradually attracted attention. These systems collect animal behavior video data through cameras and use image processing algorithms for analysis and identification, greatly improving the efficiency and accuracy of behavior monitoring.
[0003] In the prior art, common rat behavior monitoring systems have the following main deficiencies:
[0004] Limitations of closed cages: Closed cages restrict the free movement of rats and may affect their natural behavior. In addition, the closed environment also increases the difficulty of video acquisition and is prone to poor video quality.
[0005] Insufficient video capture quality: Many existing systems use low-resolution cameras, resulting in unclear video data being collected, affecting the accuracy of subsequent behavior recognition.
[0006] Limitations of behavior recognition algorithms: Traditional behavior recognition algorithms mostly rely on simple image processing technology and cannot accurately identify complex behavior patterns, especially when faced with detailed analysis such as facial expression recognition, the recognition effect is poor.
[0007] Lack of physiological parameter monitoring: In the existing technology, most systems only focus on the external manifestations of behavior and lack real-time monitoring of rat physiological parameters such as heart rate and body temperature, which limits the comprehensive assessment of the health status of rats.
[0008] There is a lack of solutions for effectively compressing and storing large amounts of video data and physiological parameter data. In addition, existing technologies also have certain limitations in terms of efficient encoding and compression of video data, making it difficult to balance data quality and storage efficiency. Summary of the invention
[0009] In view of the limitations of closed cages, insufficient video acquisition quality, limitations of behavior recognition algorithms, and lack of physiological parameter monitoring, the present invention proposes a rat behavior monitoring system and method based on transparent cages and image processing. By using cages made of high-strength transparent materials, not only the freedom of rat activities is guaranteed, but also video acquisition is not disturbed. At the same time, the use of high-definition cameras and advanced image processing algorithms improves the accuracy and reliability of behavior recognition. In addition, the system also integrates a non-contact physiological parameter monitoring module to achieve real-time monitoring of physiological parameters such as rat heart rate and body temperature, thereby providing a more comprehensive health assessment.
[0010] In order to achieve the above object, the present invention is implemented by the following technical solutions:
[0011] A rat behavior monitoring system based on transparent cages and image processing, the system comprising:
[0012] Transparent cages for containing and restricting the activities of rats;
[0013] High-definition cameras, arranged around the transparent cages, were used to collect video data of rat behavior;
[0014] A video acquisition module, used for receiving and preprocessing video data acquired by the high-definition camera;
[0015] Behavior recognition module, used to process the preprocessed video data based on the DeepBehaviorNet algorithm model to identify the behavior of rats;
[0016] A facial expression recognition module, for analyzing facial features of rats in the video data to assess pain and emotional states of the rats;
[0017] A physiological parameter monitoring module is used to collect the physiological parameters of rats in a non-contact manner;
[0018] Data storage and analysis module, which stores and analyzes identification and monitoring results;
[0019] Behavior prediction and anomaly detection module, which predicts rat behavior and detects anomalies based on time series models;
[0020] Data analysis and report generation module, providing data visualization and experimental report generation;
[0021] Remote monitoring and collaboration module, which realizes remote monitoring and multi-user collaboration based on cloud platform.
[0022] As a preferred solution of the present invention, the transparent cage is made of high-strength transparent material, and the structural design allows the rat's activities to be unrestricted and the video collection to be undisturbed.
[0023] As a preferred solution of the present invention, the high-definition camera has a resolution of at least 1080P and a frame rate of 60 frames per second to ensure high-quality video data acquisition.
[0024] As a preferred solution of the present invention, the behavior recognition module includes a DeepBehaviorNet algorithm model that combines a deep convolutional neural network and a long short-term memory network, which is used to extract spatial features from video data and combine time series analysis to identify complex behavior patterns. The specific steps include:
[0025] Input the video frames into a deep convolutional neural network to extract spatial features;
[0026] The extracted spatial features are weighted using an attention mechanism to highlight key local facial areas;
[0027] Inputting the weighted features into a long short-term memory network to perform time series analysis to identify the specific behavior and pain state of the rat;
[0028] The final behavior recognition and pain assessment are performed by fusing multimodal features including facial expression features, body posture features and physiological parameter features.
[0029] As a preferred solution of the present invention, the facial expression recognition module adopts an improved multimodal fusion algorithm to fuse facial expression features with body posture features, and realizes emotion assessment through the following steps:
[0030] a: Preprocess the video data, including denoising, contrast enhancement and standardization, to ensure the quality and consistency of the input data;
[0031] b: extracting facial expression features and body posture features from the preprocessed image using a convolutional neural network;
[0032] c: using an attention mechanism to weight the extracted facial expression features and body posture features to highlight key facial local areas and body posture features;
[0033] The attention mechanism focuses more on key facial local areas and body posture features by calculating the importance weight (α) of each feature. The formula is as follows:
[0034] The weight calculation formula is:
[0035]
[0036] Among them, e i represents the importance score of feature i, α iis the weight of feature i. The attention mechanism ensures that important features are given higher weights when fused, thus improving recognition accuracy.
[0037] d: fusing the weighted facial expression features with the body posture features to generate a comprehensive feature vector;
[0038] e: The emotional state of the rats was assessed by the following formula:
[0039]
[0040] Among them, E t represents the emotion evaluation value at time t, F t Indicates facial expression features, B t Represents body posture characteristics, P t Represents physiological parameter characteristics, α i ,β i ,γ i is the weight parameter of the current time feature, which is tuned through training data, and N is the size of the time window.
[0041] As a preferred solution of the present invention, the physiological parameter monitoring module includes a non-contact infrared sensor for real-time monitoring of the heart rate and body temperature of the rat, and data processing is achieved through the following steps:
[0042] Step 1: Use a non-contact infrared sensor to collect the rat's heart rate and body temperature data in real time;
[0043] Step 2: Preprocess the collected data, including filtering, denoising and standardization, to ensure the accuracy and consistency of the data;
[0044] Step 3: Extract key features from the preprocessed data, including heart rate waveform features and body temperature change features;
[0045] Step 4: Use the weighted average method to fuse the heart rate and body temperature data collected multiple times to generate comprehensive physiological parameters;
[0046] Step 5: Process the heart rate data using the following formula:
[0047]
[0048] Among them, H t represents the heart rate at time t, I i represents the image data collected by the infrared sensor, T t-j represents the body temperature data at time tj, w i is the image data weight coefficient within the time window, v j is the weight coefficient of body temperature data within the time window, U kis the weighting coefficient for combining image data and temperature data, N is the size of the image data time window, M is the size of the temperature data time window, and L is the size of the combined data time window;
[0049] Step 6: Input the processed heart rate and body temperature data into the behavior recognition module and emotion assessment module to assist in identifying the behavior and emotional state of the rats.
[0050] As a preferred solution of the present invention, the data storage and analysis module includes a database system for storing video data, physiological parameters and recognition results, and generating a multi-dimensional behavior analysis report through a data analysis tool.
[0051] As a preferred solution of the present invention, the data storage and analysis module further includes a data compression algorithm module for compressing a large amount of stored video data and physiological parameter data without affecting the data quality.
[0052] As a preferred solution of the present invention, the data compression algorithm includes the following steps: dividing the input video into coding tree units, predicting, transforming, quantizing, encoding and adaptive filtering.
[0053] As a preferred solution of the present invention, the behavior prediction and anomaly detection module is based on an improved long short-term memory network and performs behavior prediction through the following formula:
[0054]
[0055] Among them, P(t+1) is the behavior prediction value at time t, σ is the activation function, and W p is the weight matrix, h t is the hidden state at time t, b p is the bias, η is the adjustment coefficient, δ i is the weight of the historical state, k is the window size of the historical state, and is implemented through the following steps:
[0056] Extract historical behavior data from the database and construct time series;
[0057] Inputting the time series into an improved long short-term memory network model, and using the weight adjustment of historical states to predict future behavior patterns;
[0058] Perform anomaly detection on the prediction results to identify abnormal behaviors.
[0059] As a preferred embodiment of the present invention, the data analysis and report generation module provides an automatic report generation function based on machine learning, including behavioral statistics, abnormal records and health assessment. The generated report content includes but is not limited to: overall behavioral statistics of rats, behavioral change trend analysis, abnormal behavior records and cause analysis, health assessment and suggestions.
[0060] A rat behavior monitoring method based on transparent cages and image processing, the method comprising:
[0061] Step S1: placing the rat in a transparent cage and using a high-definition camera to collect video data of the rat's behavior;
[0062] Step S2: preprocessing the video data, including denoising, contrast enhancement and standardization, to ensure data quality and consistency;
[0063] Step S3: input the preprocessed video data into the DeepBehaviorNet algorithm model that combines a deep convolutional neural network and a long short-term memory network, extract spatial features and perform time series analysis to identify the behavior and pain state of the rats;
[0064] Step S4: using the facial expression recognition module to extract the facial expression features and body posture features of the rat from the video data, and weighting the key features through the attention mechanism;
[0065] Step S5: fusing the weighted facial expression features and body posture features with the physiological parameter features to generate a comprehensive feature vector for evaluating the emotional state of the rat;
[0066] Step S6: using a non-contact infrared sensor to monitor the heart rate and body temperature of the rat in real time, preprocessing and feature extraction of the collected data, and combining the video data for behavior recognition and pain assessment;
[0067] Step S7: storing the processed data in the data storage and analysis module, and generating a multi-dimensional behavior analysis report through a data analysis tool;
[0068] Step S8: predicting the future behavior pattern of the rat based on the time series model, detecting abnormal behavior, and generating a prediction and abnormality detection report;
[0069] Step S9: Through the remote monitoring and collaboration module, remote real-time monitoring of rat behavior and multi-user collaboration are achieved.
[0070] Compared with the prior art, the beneficial effects of the present invention are as follows: the transparent cage of the present invention is combined with a high-definition camera, so that the video data is clear and unobstructed, and the accuracy of behavior recognition is improved. The design of the transparent cage avoids the visual obstruction of the traditional cage, so that the algorithm can more accurately capture the behavioral characteristics of the rat. The DeepBehaviorNet algorithm model combining the deep convolutional neural network and the long short-term memory network can accurately identify the complex behavior patterns of the rat by extracting the spatial features in the video data and performing time series analysis. Compared with the traditional image processing method, the algorithm can more accurately identify the specific behavior and pain state of the rat. The introduction of the multimodal fusion algorithm and the attention mechanism enables the system to have higher accuracy in identifying complex behaviors. By fusing multiple features and dynamically adjusting the importance of each feature, the system can more accurately identify the facial expressions, body postures and physiological states of the rat. The attention mechanism ensures that the key features are given higher weights when fused, improving the recognition accuracy. The application of non-contact infrared sensors realizes real-time monitoring of heart rate and body temperature, and provides comprehensive assessment data of the health status of rats. Combined with behavioral recognition data, researchers can have a more comprehensive understanding of the health and behavior of rats, providing more accurate data support for experimental research. The data storage and analysis module can store video data, physiological parameters and recognition results, and generate multi-dimensional behavioral analysis reports through data analysis tools. It not only improves the efficiency of data management, but also facilitates the generation of experimental reports. The efficient video encoding method ensures that the quality of the data is not affected while compressing the video data, providing technical support for long-term storage and big data analysis. Based on the improved long short-term memory network, the time series model is used to predict rat behavior and detect anomalies. It can not only predict the future behavior patterns of rats, but also identify abnormal behaviors in time, providing an early warning mechanism for research. Remote monitoring and multi-user collaboration are achieved through the cloud platform, allowing researchers to monitor the progress of the experiment and collaborate anytime and anywhere, improving the flexibility and efficiency of experimental management. Remote monitoring and multi-user collaboration functions can significantly improve the efficiency and effectiveness of research work. BRIEF DESCRIPTION OF THE DRAWINGS
[0071] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.
[0072] in:
[0073] Figure 1 is an overall structural block diagram of the system in an embodiment of the present invention;
[0074] Figure 2is a schematic diagram of the implementation steps of the behavior recognition module in an embodiment of the present invention;
[0075] Figure 3 Schematic diagram of the implementation steps of the facial expression recognition module in an embodiment of the present invention;
[0076] Figure 4 is a schematic diagram of the data processing flow of the physiological parameter monitoring module in an embodiment of the present invention;
[0077] Figure 5 is a flow chart of a method according to an embodiment of the present invention;
[0078] Figure 6 It is a facial key point and position marking diagram in the rat behavior monitoring system experiment of the present invention;
[0079] Figure 7 It is a schematic diagram of pain detection in a rat behavior monitoring system experiment of the present invention;
[0080] Figure 8 It is a schematic diagram of detecting the pain-free state in a rat behavior monitoring system experiment of the present invention. DETAILED DESCRIPTION
[0081] In order to make the purpose, technical solution and advantages of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the described embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.
[0082] Example 1
[0083] like Figure 1 As shown, it is an embodiment of the present invention, which provides a rat behavior monitoring system based on transparent cages and image processing, and the system includes:
[0084] Transparent cages for containing and restricting the activities of rats;
[0085] Furthermore, the transparent cage is made of high-strength transparent material, and the structural design allows the rat's activities to be unrestricted and the video collection to be undisturbed.
[0086] Transparent cage: used to contain and restrict the activities of rats, made of high-strength transparent material, to ensure that the rats' activities are not restricted and the video collection is not disturbed. It helps to avoid interference of external factors on video collection, while ensuring that the rats move in a natural state.
[0087] In one embodiment, high-definition cameras are arranged around the transparent cage to collect video data of rat behavior;
[0088] Furthermore, the high-definition camera has a resolution of at least 1080P and a frame rate of 60 frames per second to ensure high-quality video data acquisition.
[0089] HD cameras: They are placed around the transparent cages, with a resolution of at least 1080P and a frame rate of 60 frames per second, to collect video data of rat behavior. HD cameras can capture clear image details and provide high-quality data input for subsequent image processing and behavior recognition.
[0090] The transparent cages are made of highly transparent materials such as polycarbonate to ensure optical clarity, impact resistance and durability. The position of the HD camera is precisely calculated to ensure that all angles of the rat's activities can be captured. The combination of transparent cages and HD cameras can obtain clear and interference-free behavioral video data.
[0091] In one of the embodiments, a video acquisition module is used to receive and pre-process the video data acquired by the high-definition camera;
[0092] Video acquisition module: used to receive and preprocess video data collected by high-definition cameras. This module includes preprocessing steps such as denoising, contrast enhancement, and standardization to ensure the quality and consistency of input data. These preprocessing steps can improve the accuracy and stability of subsequent algorithms.
[0093] In a specific embodiment, Figure 2 As shown, a behavior recognition module is used to process the preprocessed video data based on the DeepBehaviorNet algorithm model to recognize the behavior of rats;
[0094] Furthermore, the behavior recognition module includes a DeepBehaviorNet algorithm model that combines a deep convolutional neural network and a long short-term memory network, which is used to extract spatial features from video data and combine time series analysis to identify complex behavior patterns. The specific steps include:
[0095] Input the video frames into a deep convolutional neural network to extract spatial features;
[0096] The extracted spatial features are weighted using an attention mechanism to highlight key local facial areas;
[0097] Inputting the weighted features into a long short-term memory network to perform time series analysis to identify the specific behavior and pain state of the rat;
[0098] The final behavior recognition and pain assessment are performed by fusing multimodal features including facial expression features, body posture features and physiological parameter features.
[0099] Facial expression feature extraction uses a deep learning network, a convolutional neural network, to capture subtle changes in the rat's face. Body posture features use key point detection technology to track the posture changes of the rat in real time. Physiological parameter features include heart rate and body temperature, which are acquired using non-contact infrared sensors. Through the attention mechanism, the weights are dynamically adjusted according to the importance of different features to improve the accuracy of overall recognition.
[0100] Behavior recognition module: The pre-processed video data is processed based on the DeepBehaviorNet algorithm model to identify the behavior of rats. The DeepBehaviorNet algorithm model combines deep convolutional neural networks and long short-term memory networks to realize the recognition of complex behavior patterns of rats by extracting spatial features, weighted processing and time series analysis.
[0101] In a specific embodiment, Figure 3 As shown, a facial expression recognition module is used to analyze the facial features of rats in the video data to evaluate the pain and emotional state of the rats;
[0102] Furthermore, the facial expression recognition module adopts an improved multimodal fusion algorithm to fuse facial expression features with body posture features, and realizes emotion evaluation through the following steps:
[0103] a: Preprocess the video data, including denoising, contrast enhancement and standardization, to ensure the quality and consistency of the input data;
[0104] b: extracting facial expression features and body posture features from the preprocessed image using a convolutional neural network;
[0105] c: using an attention mechanism to weight the extracted facial expression features and body posture features to highlight key facial local areas and body posture features;
[0106] d: fusing the weighted facial expression features with the body posture features to generate a comprehensive feature vector;
[0107] e: The emotional state of the rats was assessed by the following formula:
[0108]
[0109] Among them, E t represents the emotion evaluation value at time t, F t Indicates facial expression features, B t Represents body posture characteristics, Pt Represents physiological parameter characteristics, α i ,β i ,γ i is the weight parameter of the current time feature, which is tuned through training data, and N is the size of the time window.
[0110] Facial expression recognition module: used to analyze the facial features of rats in video data to assess the pain and emotional state of rats. This module uses a multimodal fusion algorithm to extract facial expression features and body posture features, and uses an attention mechanism for weighted processing to ultimately generate a comprehensive feature vector for emotion assessment.
[0111] In one embodiment, if Figure 4 As shown, the physiological parameter monitoring module is used to collect the physiological parameters of rats in a non-contact manner;
[0112] Furthermore, the physiological parameter monitoring module includes a non-contact infrared sensor for real-time monitoring of the heart rate and body temperature of the rat, and data processing is achieved through the following steps:
[0113] Step 1: Use a non-contact infrared sensor to collect the rat's heart rate and body temperature data in real time;
[0114] Step 2: Preprocess the collected data, including filtering, denoising and standardization, to ensure the accuracy and consistency of the data;
[0115] Step 3: Extract key features from the preprocessed data, including heart rate waveform features and body temperature change features;
[0116] Step 4: Use the weighted average method to fuse the heart rate and body temperature data collected multiple times to generate comprehensive physiological parameters;
[0117] Step 5: Process the heart rate data using the following formula:
[0118]
[0119] Wherein, Ht represents the heart rate at time t, Ii represents the image data collected by the infrared sensor, Tt-j represents the body temperature data at time tj, wi is the weight coefficient of the image data in the time window, vj is the weight coefficient of the body temperature data in the time window, uk is the weight coefficient of combining image data and body temperature data, N is the size of the image data time window, M is the size of the body temperature data time window, and L is the size of the combined data time window;
[0120] Step 6: Input the processed heart rate and body temperature data into the behavior recognition module and emotion assessment module to assist in identifying the behavior and emotional state of the rats.
[0121] Physiological parameter monitoring module: It includes a non-contact infrared sensor for real-time collection of rats’ heart rate and body temperature. This module extracts key features and generates comprehensive physiological parameters by filtering, denoising and standardizing the collected data to assist in behavior and emotion recognition.
[0122] In one embodiment, a data storage and analysis module stores and analyzes identification and monitoring results;
[0123] Furthermore, the data storage and analysis module includes a database system for storing video data, physiological parameters and recognition results, and generating a multi-dimensional behavior analysis report through a data analysis tool.
[0124] Data storage and analysis module: Store and analyze recognition and monitoring results, and generate multi-dimensional behavior analysis reports. This module includes a database system for storing video data, physiological parameters, and recognition results, and generates multi-dimensional behavior analysis reports through data analysis tools to provide detailed data support for researchers.
[0125] Furthermore, the data storage and analysis module further includes a data compression algorithm module for compressing a large amount of stored video data and physiological parameter data without affecting the data quality.
[0126] The data compression algorithm includes the following steps: dividing the input video into coding tree units, predicting, transforming, quantizing, encoding and adaptive filtering.
[0127] The data compression algorithm is a high-efficiency video encoding method, comprising the following steps:
[0128] Dividing an input video frame into coding tree units, each coding tree unit includes a plurality of coding blocks, wherein the coding blocks are further divided into prediction blocks and transform blocks;
[0129] Using the coded pixels in the current frame to predict the uncoded pixels, including prediction modes in multiple directions, and selecting the optimal prediction mode;
[0130] The pixels in the reference frame are used to predict the pixels in the current frame, and the best matching block is found through motion estimation and compensation. Motion estimation uses a block matching algorithm, and motion compensation uses motion vectors with integer and sub-pixel accuracy.
[0131] Perform discrete cosine transform on the prediction residuals;
[0132] quantizing the transformed coefficients using an adaptive quantization method based on quantization parameters;
[0133] Adaptive arithmetic coding through context information;
[0134] Encoding is performed by a variable length coding method;
[0135] Adaptively filter the encoded block boundaries to reduce the blocking effect, and the filtering parameters are dynamically adjusted according to the texture characteristics of the block boundaries;
[0136] Applied after deblocking filtering, it further reduces compression-induced distortion through classification and compensation techniques. Classification is based on the statistical characteristics of pixel values, and compensation includes edge compensation and bandpass compensation.
[0137] The compressed and encoded data information is organized into a bit stream that complies with the HEVC standard, including video data, parameter sets, and reference frame information. The bit stream adopts a hierarchical structure to support flexible decoding and random access.
[0138] In one embodiment, the behavior prediction and anomaly detection module predicts rat behavior and detects anomalies based on a time series model;
[0139] Furthermore, the behavior prediction and anomaly detection module is based on an improved long short-term memory network and performs behavior prediction through the following formula:
[0140]
[0141] Among them, P(t+1) is the behavior prediction value at time t, σ is the activation function, Wp is the weight matrix, ht is the hidden state at time t, bp is the bias, η is the adjustment coefficient, δi is the weight of the historical state, k is the window size of the historical state, and it is implemented by the following steps:
[0142] Extract historical behavior data from the database and construct time series;
[0143] Inputting the time series into an improved long short-term memory network model, and using the weight adjustment of historical states to predict future behavior patterns;
[0144] Perform anomaly detection on the prediction results to identify abnormal behaviors.
[0145] Behavior prediction and anomaly detection module: predict rat behavior and detect anomalies based on the time series model. This module extracts historical behavior data from the database, constructs a time series, and inputs it into the improved long short-term memory network model. It uses the weight adjustment of historical states to predict future behavior patterns and detect anomalies on the predicted results.
[0146] In one of the embodiments, the data analysis and report generation module provides data visualization and experimental report generation;
[0147] Furthermore, the data analysis and report generation module provides an automatic report generation function based on machine learning, including behavioral statistics, abnormal records and health assessments. The generated report content includes but is not limited to: overall behavioral statistics of rats, behavioral change trend analysis, abnormal behavior records and cause analysis, health assessments and recommendations.
[0148] Data analysis and report generation module: provides data visualization and experimental report generation. This module is based on the automatic report generation function of machine learning, including behavior statistics, abnormal records and health assessment. The generated report content includes but is not limited to the overall behavior statistics of rats, behavior change trend analysis, abnormal behavior records and cause analysis, health assessment and suggestions.
[0149] Furthermore, the remote monitoring and collaboration module realizes remote monitoring and multi-user collaboration based on the cloud platform.
[0150] Remote monitoring and collaboration module: Remote monitoring and multi-user collaboration are realized based on the cloud platform. Through this module, researchers can realize remote real-time monitoring of rat behavior and support multi-user collaboration, which improves the flexibility and practicality of the system.
[0151] Explanation of terms
[0152] DeepBehaviorNet: An algorithmic model that combines deep convolutional neural networks and long short-term memory networks to extract spatial features from video data and combine them with time series analysis to identify complex behavioral patterns.
[0153] Multimodal fusion algorithm: an algorithm that combines multiple data sources (such as facial expression features, body posture features, and physiological parameter features) for analysis to improve recognition accuracy.
[0154] Attention mechanism: A technique for highlighting key features by calculating the importance weight of each feature and focusing more attention on key facial local areas and body posture features.
[0155] Non-contact infrared sensor: FLIR sensor, which monitors heart rate and body temperature by capturing infrared radiation. The sensor is installed on the top or side of the transparent cage to ensure that it does not affect the free movement of the rats and can obtain data in real time. The sensor data is filtered out of noise through signal processing algorithms to ensure the accuracy and consistency of monitoring.
[0156] Example 2
[0157] like Figure 5 FIG. 1 is another embodiment of the present invention, which provides a rat behavior monitoring method based on a transparent cage and image processing, and the method comprises:
[0158] Step S1: placing the rat in a transparent cage and using a high-definition camera to collect video data of the rat's behavior;
[0159] Step S2: preprocessing the video data, including denoising, contrast enhancement and standardization, to ensure data quality and consistency;
[0160] Step S3: input the preprocessed video data into the DeepBehaviorNet algorithm model that combines a deep convolutional neural network and a long short-term memory network, extract spatial features and perform time series analysis to identify the behavior and pain state of the rats;
[0161] Step S4: using the facial expression recognition module to extract the facial expression features and body posture features of the rat from the video data, and weighting the key features through the attention mechanism;
[0162] Step S5: fusing the weighted facial expression features and body posture features with the physiological parameter features to generate a comprehensive feature vector for evaluating the emotional state of the rat;
[0163] Step S6: using a non-contact infrared sensor to monitor the heart rate and body temperature of the rat in real time, preprocessing and feature extraction of the collected data, and combining the video data for behavior recognition and pain assessment;
[0164] Step S7: storing the processed data in the data storage and analysis module, and generating a multi-dimensional behavior analysis report through a data analysis tool;
[0165] Step S8: predicting the future behavior pattern of the rat based on the time series model, detecting abnormal behavior, and generating a prediction and abnormality detection report;
[0166] Step S9: Through the remote monitoring and collaboration module, remote real-time monitoring of rat behavior and multi-user collaboration are achieved.
[0167] Example 3
[0168] In another embodiment of the present invention, in order to verify the monitoring system of the present invention, we designed a series of comparative experiments, and verified the performance advantages of the system of the present invention in multiple key indicators by conducting comparative experiments on 20 healthy rats. The experiment was divided into two groups, each with 10 rats, and the traditional closed cage video monitoring system and the transparent cage and image processing system of the present invention were used for behavior monitoring and data collection.
[0169] Experimental steps:
[0170] Experimental preparation: 20 rats were randomly divided into two groups, A and B, with 10 rats in each group. Group A used a traditional closed cage video monitoring system, and Group B used a transparent cage and image processing system of the present invention.
[0171] Data collection: In the monitoring area, rats in group A and group B were placed in corresponding cages separately, and the experimental time for each rat was 24 hours. A high-definition camera was used to continuously record the behavioral video data of the rats, and a non-contact infrared sensor was used to monitor the heart rate and body temperature data of the rats in real time.
[0172] Data processing: Video data is collected by high-definition cameras and stored in the data storage and analysis system. The behavior recognition module processes the collected video data and identifies the behavior patterns of rats. The facial expression recognition module recognizes facial expressions in the video data and evaluates the emotional state of rats. The physiological parameter monitoring module preprocesses and analyzes the real-time collected heart rate and body temperature data to generate comprehensive physiological parameters.
[0173] Data analysis: All data were statistically analyzed to calculate the behavior recognition accuracy, video data clarity score, physiological parameter monitoring accuracy, video data acquisition stability score, and facial expression recognition accuracy.
[0174] Experimental comparison data
[0175] index Traditional closed cage system System of the present invention Behavior recognition accuracy 85% 95% Video data clarity rating 3.5 4.8 Physiological parameter monitoring accuracy 80% 92% Video data acquisition stability score 3 4.7 Facial expression recognition accuracy 70% 90%
[0176] Experimental results:
[0177] Behavior recognition accuracy: The behavior recognition accuracy of the traditional closed cage system is 85%, while the system of the present invention achieves a high accuracy recognition of 95% due to the use of cages made of high-strength transparent materials, combined with high-definition cameras and deep learning algorithms.
[0178] Video data clarity rating: The traditional closed cage system is rated 3.5, which is mainly affected by the cage material and structure, resulting in unclear video data. The system of the present invention benefits from the combination of transparent cage and high-definition camera, and the video data clarity rating is as high as 4.8.
[0179] Physiological parameter monitoring accuracy: Due to technical limitations, the accuracy of traditional systems is 80%. The system of the present invention adopts non-contact infrared sensors, combined with data preprocessing and fusion analysis technology, to achieve an accuracy of 92%, greatly improving the monitoring accuracy.
[0180] Video data acquisition stability score: The traditional system scored 3.0, mainly due to cage occlusion and reflection problems, which led to unstable video acquisition. The system of the present invention scored 4.7, proving the significant advantage of the transparent cage design in acquisition stability.
[0181] Facial expression recognition accuracy: Due to the quality of video data, the recognition accuracy of traditional systems is only 70%. However, the system of the present invention achieves a high recognition accuracy of 90% through an improved multimodal fusion algorithm, combined with a high-definition camera and a transparent cage design.
[0182] Experimental data summary:
[0183] It can be seen from the above experimental comparison data that the rat behavior monitoring system based on transparent cages and image processing of the present invention has shown significant superiority in multiple key indicators, especially in terms of behavior recognition accuracy, video data quality and physiological parameter monitoring accuracy, which is significantly better than the traditional closed cage system. The experimental results fully demonstrate the beneficial effects of the system of the present invention and provide reliable technical support for rat behavior research and health monitoring.
[0184] like Figure 6 As shown, the facial key points and position marks of a rat are shown. The red box in the figure marks the main positions of the rat's face, and the digital labels mark the specific positions (such as No. 2 and No. 5 are located on the ears). The blue digital labels (1-18) mark the facial key points, which are used to describe the various parts of the face in detail. The yellow area shows the pain level label and AU label, specifically:
[0185] Pain level label: 1
[0186] AU tags: eyes:1, nose:0, ears:1, whiskers:0.
[0187] like Figure 7 The figure shows a rat in an experimental environment with its face marked as in a state of pain. The red box marks the face of the rat, and the label is "pain 0.91", indicating that the system detects that the rat has a probability of pain of 0.91. This is used to demonstrate the practical application of the system in rat facial expression recognition.
[0188] like Figure 8 The figure shows the detection result of a rat in profile. The orange frame marks the face of the rat, and the label is "nopain 0.89", indicating that the system detects that the rat has a probability of 0.89 of no pain. This figure is used to demonstrate the system's ability to monitor the behavior and expressions of rats at different angles and states.
[0189] In summary, the behavior recognition accuracy of the system of the present invention reached 95%, which is 10 percentage points higher than the 85% of the traditional closed cage system. This is due to the transparent cage design and the behavior recognition module combined with the deep learning algorithm, which effectively reduces external interference and improves the accuracy of recognition. The use of a high-definition camera (1080P, 60 frames per second) ensures the high resolution and high frame rate of the video data, and the video data clarity score reaches 4.8 (five-point system), which is much higher than the 3.5 of the traditional system. The use of transparent cages eliminates the problems of reflection and occlusion, making the video data clearer and more detailed. The non-contact infrared sensor is integrated to monitor the heart rate and body temperature of rats in real time, and the accuracy of physiological parameter monitoring reaches 92%, which is significantly better than the 80% of the traditional system. This non-invasive monitoring method avoids direct interference with rats and improves the accuracy and reliability of the data. The efficient video encoding method ensures that the quality of the data is not affected while compressing the video data, providing technical support for long-term storage and big data analysis. Through experimental comparison, the video data acquisition stability score of the system of the present invention is 4.7, which is significantly higher than the 3.0 of the traditional system. This shows that the transparent cage design plays an important role in reducing reflections and occlusions, ensuring the stability and continuity of video acquisition. The improved multimodal fusion algorithm is used to comprehensively analyze facial expression features and body posture features, and the accuracy of facial expression recognition reaches 90%, which is much higher than the 70% of the traditional system. This makes the assessment of the emotional state of rats more accurate and provides more comprehensive data support for behavioral research.
[0190] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.
[0191] Any process or method description in the flow chart or otherwise described herein can be understood to represent a module, fragment or portion of a code including one or more executable instructions for implementing the steps of a specific logical function or process. And the scope of the preferred embodiment of the present application includes other implementations, in which the functions may not be performed in the order shown or discussed, including in a substantially simultaneous manner or in a reverse order according to the functions involved.
[0192] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.
Claims
1. A rat behavior monitoring system based on transparent cages and image processing, characterized in that: include: Transparent cages for containing and restricting the activities of rats; High-definition cameras, arranged around the transparent cages, were used to collect video data of rat behavior; A video acquisition module, used for receiving and preprocessing video data acquired by the high-definition camera; Behavior recognition module, used to process the preprocessed video data based on the DeepBehaviorNet algorithm model to identify the behavior of rats; A facial expression recognition module, for analyzing facial features of rats in the video data to assess pain and emotional states of the rats; A physiological parameter monitoring module is used to collect the physiological parameters of rats in a non-contact manner; Data storage and analysis module, which stores and analyzes identification and monitoring results; Behavior prediction and anomaly detection module, which predicts rat behavior and detects anomalies based on time series models; Data analysis and report generation module, providing data visualization and experimental report generation; Remote monitoring and collaboration module, which realizes remote monitoring and multi-user collaboration based on cloud platform.
2. The rat behavior monitoring system based on transparent cage and image processing according to claim 1, characterized in that: The transparent cage is made of high-strength transparent material, and the structure is designed so that the rat's activities are not restricted and the video collection is not disturbed.
3. The rat behavior monitoring system based on transparent cage and image processing according to claim 1, characterized in that: The high-definition camera has a resolution of at least 1080P and a frame rate of 60 frames per second.
4. The rat behavior monitoring system based on transparent cage and image processing according to claim 1, characterized in that: The behavior recognition module includes a DeepBehaviorNet algorithm model that combines a deep convolutional neural network and a long short-term memory network, which is used to extract spatial features from video data and combine time series analysis to identify complex behavior patterns. The specific steps include: Input the video frames into a deep convolutional neural network to extract spatial features; The extracted spatial features are weighted using an attention mechanism to highlight key local facial areas; Inputting the weighted features into a long short-term memory network to perform time series analysis to identify the specific behavior and pain state of the rat; The final behavior recognition and pain assessment are performed by fusing multimodal features including facial expression features, body posture features and physiological parameter features.
5. The rat behavior monitoring system based on transparent cage and image processing according to claim 1, characterized in that: The facial expression recognition module adopts an improved multimodal fusion algorithm to fuse facial expression features with body posture features, and realizes emotion evaluation through the following steps: a: Preprocess the video data, including denoising, contrast enhancement and standardization; b: extracting facial expression features and body posture features from the preprocessed image using a convolutional neural network; c: using the attention mechanism to perform weighted processing on the extracted facial expression features and body posture features, highlighting the key facial local areas and body posture features; d: fusing the weighted facial expression features with the body posture features to generate a comprehensive feature vector; e: The emotional state of the rats was assessed by the following formula: Among them, E t represents the emotion evaluation value at time t, F t Indicates facial expression features, B t Represents body posture characteristics, P t Represents physiological parameter characteristics, α i ,β i ,γ i is the weight parameter of the current time feature, which is tuned through training data, and N is the size of the time window.
6. The rat behavior monitoring system based on transparent cage and image processing according to claim 1, characterized in that: The physiological parameter monitoring module includes a non-contact infrared sensor for real-time monitoring of the heart rate and body temperature of the rat, and realizes data processing through the following steps: Step 1: Use a non-contact infrared sensor to collect the rat's heart rate and body temperature data in real time; Step 2: Preprocess the collected data, including filtering, denoising and standardization; Step 3: Extract key features from the preprocessed data, including heart rate waveform features and body temperature change features; Step 4: Use the weighted average method to fuse the heart rate and body temperature data collected multiple times to generate comprehensive physiological parameters; Step 5: Process the heart rate data using the following formula: Among them, H t represents the heart rate at time t, I i represents the image data collected by the infrared sensor, T t-j represents the body temperature data at time tj, w i is the image data weight coefficient within the time window, v j is the weight coefficient of body temperature data in the time window, u k is the weighting coefficient for combining image data and body temperature data, N is the size of the image data time window, M is the size of the temperature data time window, and L is the size of the combined data time window; Step 6: Input the processed heart rate and body temperature data into the behavior recognition module and emotion assessment module to assist in identifying the behavior and emotional state of the rats.
7. The rat behavior monitoring system based on transparent cage and image processing according to claim 1, characterized in that: The data storage and analysis module includes a database system for storing video data, physiological parameters and recognition results, and generating a multi-dimensional behavior analysis report through a data analysis tool.
8. The rat behavior monitoring system based on transparent cage and image processing according to claim 7, characterized in that: The data storage and analysis module further includes a data compression algorithm module for compressing a large amount of stored video data and physiological parameter data without affecting the data quality.
9. The rat behavior monitoring system based on transparent cage and image processing according to claim 8, characterized in that: The data compression algorithm comprises the following steps: dividing the input video into coding tree units, predicting, transforming, quantizing, encoding and adaptive filtering.
10. The rat behavior monitoring system based on transparent cage and image processing according to claim 1, characterized in that: The behavior prediction and anomaly detection module is based on an improved long short-term memory network and performs behavior prediction through the following formula: Among them, P(t+1) is the behavior prediction value at time t, σ is the activation function, and W p is the weight matrix, h t is the hidden state at time t, b p is the bias, η is the adjustment coefficient, δ i is the weight of the historical state, k is the window size of the historical state, and is implemented by the following steps: Extract historical behavior data from the database and construct time series; Inputting the time series into an improved long short-term memory network model, and using the weight adjustment of historical states to predict future behavior patterns; Perform anomaly detection on the prediction results to identify abnormal behaviors.
11. The rat behavior monitoring system based on transparent cage and image processing according to claim 1, characterized in that: The data analysis and report generation module provides an automatic report generation function based on machine learning, including behavioral statistics, abnormal records and health assessment. The generated report content includes but is not limited to: overall behavioral statistics of rats, behavioral change trend analysis, abnormal behavior records and cause analysis, health assessment and suggestions.
12. A rat behavior monitoring method based on a transparent cage and image processing according to any one of claims 1 to 11, characterized in that: The method comprises: Step S1: placing the rat in a transparent cage and using a high-definition camera to collect video data of the rat's behavior; Step S2: preprocessing the video data, including denoising, contrast enhancement and standardization; Step S3: input the preprocessed video data into the DeepBehaviorNet algorithm model that combines a deep convolutional neural network and a long short-term memory network, extract spatial features and perform time series analysis to identify the behavior and pain state of the rats; Step S4: using the facial expression recognition module to extract the facial expression features and body posture features of the rat from the video data, and weighting the key features through the attention mechanism; Step S5: fusing the weighted facial expression features and body posture features with the physiological parameter features to generate a comprehensive feature vector; Step S6: using a non-contact infrared sensor to monitor the heart rate and body temperature of the rat in real time, preprocessing and feature extraction of the collected data, and combining the video data for behavior recognition and pain assessment; Step S7: storing the processed data in the data storage and analysis module, and generating a multi-dimensional behavior analysis report through a data analysis tool; Step S8: predicting the future behavior pattern of the rat based on the time series model, detecting abnormal behavior, and generating a prediction and abnormality detection report; Step S9: Through the remote monitoring and collaboration module, remote real-time monitoring of rat behavior and multi-user collaboration are achieved.
Citation Information
Patent Citations
Non-contact heart rate and body temperature measuring method based on camera
CN111407245A
Human body-temperature change prediction method based on heart rate estimation and respiration estimation
CN111714094A
Patient behavior multimodal analysis and prediction system based on statistical learning
CN111920420A
Facial expression recognition method and system combined with attention mechanism
CN112418095A
Expression and posture fused bimodal teaching evaluation method and device and storage medium
CN112541529A
Cited By
Health assessment method and system for rheumatoid arthritis synovial injury DBA / 1 mouse
CN120259221A
