Visual inspection system for animal husbandry and veterinary medicine

Abstract

The invention relates to the technical field of livestock visual inspection, in particular to a livestock veterinary visual inspection system which comprises a multi-view image acquisition module, a spectral feature extraction module, a shielding detection compensation module, a behavior pattern recognition module and a real-time early warning module. Wherein the multi-view image acquisition module is used for capturing images of animals in real time from different angles; the spectral feature extraction module is used for extracting biological feature information of animals; the shielding detection compensation module is used for recognizing the shielding condition in the image and speculating the characteristics of the shielded part by using the motion information so as to carry out corresponding compensation; and the behavior pattern recognition module is used for analyzing the behavior pattern of the animal and recognizing the abnormal behavior of the animal in real time. According to the animal health monitoring system and method, through accurate biological feature recognition and abnormal behavior detection, the efficiency and accuracy of animal health monitoring are remarkably improved, so that a timely intervention basis is provided for veterinarians, and intelligent management of animal husbandry is promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of animal husbandry visual detection, in particular to a visual detection system for animal husbandry and veterinary medicine. Background Art

[0002] In recent years, the modernization of animal husbandry has promoted the demand for animal health management. Traditional monitoring methods often rely on manual observation, which makes it difficult to achieve real-time and comprehensive monitoring of animal behavior. Especially in complex environments, accurate identification of animals' biological characteristics (such as coat color and texture) and their behavior patterns becomes particularly important. The application of existing technologies in this regard usually faces occlusion problems, resulting in insufficient monitoring of animal status and affecting overall management efficiency.

[0003] There are several major problems with existing technologies in animal behavior monitoring: first, traditional biometric recognition methods are sensitive to lighting changes and occlusions, and often fail to effectively extract the true state of the animal, resulting in reduced recognition accuracy; second, existing image analysis technologies perform poorly when dealing with complex backgrounds and overlapping occlusions, making it difficult to identify abnormal behaviors in real time. In addition, the lack of an effective real-time warning mechanism makes it impossible for veterinarians to intervene in time. Therefore, a visual detection system for animal husbandry and veterinary medicine is developed to overcome the problems of biometric extraction and occlusion detection. Summary of the invention

[0004] Based on the above objectives, the present invention provides a visual inspection system for animal husbandry and veterinary medicine.

[0005] A visual inspection system for animal husbandry and veterinary medicine, comprising a multi-view image acquisition module, a spectral feature extraction module, an occlusion detection compensation module, a behavior pattern recognition module and a real-time warning module; wherein:

[0006] Multi-view image acquisition module: deploys multiple cameras to capture images of animals from different angles in real time to form an image dataset;

[0007] Spectral feature extraction module: used to process the image data set output by the multi-view image acquisition module, and use spectral feature extraction technology to analyze the reflectivity of the color and texture of the animal at different wavelengths to extract the biological feature information of the animal;

[0008] Occlusion detection and compensation module: used to receive the image data set from the multi-view image acquisition module, identify the occlusion in the image by analyzing the motion information of the animal between different frames, and use the motion information to infer the features of the occluded part to make corresponding compensation;

[0009] Behavior pattern recognition module: Based on the biometric information provided by the spectral feature extraction module and the image data compensated by the occlusion detection and compensation module, it analyzes the animal's behavior pattern and identifies the animal's abnormal behavior in real time, including anxiety, aggression or abnormal eating;

[0010] Real-time warning module: Based on the abnormal behavior recognition results of the behavior pattern recognition module, it automatically generates health warnings and notifies veterinarians to intervene.

[0011] Optionally, the multi-view image acquisition module includes a camera deployment unit, an image capture unit and a data transmission unit; wherein:

[0012] Camera deployment unit: used to deploy high-definition cameras at predetermined locations in animal activity areas; specifically, cameras are installed at the four corners of the area to form a 360-degree monitoring view; in addition, two cameras with tilted angles are set in areas where animals frequently move;

[0013] Image capture unit: used to capture images of animals in real time, and uses automatic adjustment technology to automatically adjust the exposure value and white balance according to the ambient brightness under different lighting conditions;

[0014] Data transmission unit: used to transmit the real-time captured image data to the subsequent processing module. The video encoding standard H.264 or H.265 is used in the transmission process to reduce the data transmission volume. The specific transmission is achieved through optical fiber or Wi-Fi technology.

[0015] Optionally, the spectral feature extraction module includes an image data receiving unit, a spectral analysis unit and a biometric information extraction unit; wherein:

[0016] Image data receiving unit: used to receive the image data set output from the multi-view image acquisition module, and format the received image data set to ensure the integrity and consistency of the data;

[0017] Spectral analysis unit: used to perform spectral decomposition on the received image data, analyze the reflectivity of the animal at different wavelengths, and decompose the image into multiple spectral channels, so as to extract the color and texture characteristics of the animal and generate a spectral feature map;

[0018] Biometric information extraction unit: Based on the spectral feature graph generated by the spectral analysis unit, analyze and identify the animal's biometric information, the biometric information including fur color, texture and skin condition.

[0019] Optionally, the spectrum analysis unit includes:

[0020] Spectral decomposition: The received image data is first converted to the frequency domain through Fourier transform to identify the different frequency components in the image;

[0021] Wavelength analysis: Set different wavelength ranges and process the image through filters to generate multiple spectral channels, each of which represents the reflectance characteristics of the animal at a predetermined wavelength;

[0022] Feature extraction: In the generated spectral channels, a clustering algorithm is used to extract the color and texture features of the animal, and the mean and variance of the pixels in each channel are calculated to identify the characteristic areas, thereby extracting the biological characteristic information of the animal;

[0023] Spectral feature map generation: The color and texture features extracted at different wavelengths are integrated to generate a spectral feature map;

[0024] Ambient light interference elimination: During the entire process of spectral analysis unit operation, regularization techniques are applied to eliminate ambient light interference to optimize the quality of image data.

[0025] Optionally, the occlusion detection and compensation module includes a motion information extraction unit, an occlusion detection unit, a feature inference unit and a compensation generation unit; wherein:

[0026] Motion information extraction unit: used to receive the image data set from the multi-view image acquisition module, and use the Lucas-Kanade algorithm to calculate the motion vector of the animal between different frames, identify the motion direction and speed of each pixel, and thus obtain the movement trajectory and dynamic behavior of the animal;

[0027] Occlusion detection unit: By analyzing motion information, comparing motion patterns between different frames, establishing an optical flow field, and identifying the mutation area of ​​the motion vector, the mutation area corresponds to the part occluded by other objects, so as to accurately detect the occlusion in the image;

[0028] Feature inference unit: used to infer the features of the occluded part using motion information after identifying the occlusion situation;

[0029] Compensation generation unit: used to combine the inferred features with the actual image. The generated compensated image can be seamlessly embedded into the original image, thereby restoring the complete appearance of the animal and reducing the information loss caused by occlusion.

[0030] Optionally, the motion information extraction unit includes:

[0031] Image frame acquisition: first obtain continuous image frames from the multi-view image acquisition module;

[0032] Optical flow calculation: Lucas-Kanade algorithm is used to calculate the motion vector of each pixel by analyzing the image brightness changes between adjacent frames, and identify its displacement in the x and y directions;

[0033] Least squares solution: Construct a linear equation containing multiple neighborhood pixels and use the least squares method to solve it, so as to obtain the movement direction and speed of each pixel, and then form complete motion vector data;

[0034] Movement trajectory acquisition: Using the calculated motion vector data, the animal's movement trajectory is formed through a tracking algorithm to reflect the animal's dynamic behavior in real time.

[0035] Optionally, the occlusion detection unit specifically includes:

[0036] Optical flow field construction: First, the motion vector data provided by the motion information extraction unit is used to construct the optical flow field, which is used to represent the motion direction and speed of each pixel between different frames;

[0037] Motion pattern comparison: compare the optical flow fields between adjacent frames, identify changes in motion patterns, and determine areas where motion vectors change suddenly, especially focusing on areas that are inconsistent with surrounding motion vectors;

[0038] Mutation area identification: Use the set threshold to filter the motion vector and identify the area where the motion vector changes significantly.

[0039] Optionally, the feature inference unit specifically includes:

[0040] Unoccluded area recognition: first identify the unoccluded pixel area;

[0041] Pixel value extraction: Extract pixel values ​​from unobstructed areas to provide basic data for subsequent analysis;

[0042] Mean calculation: Statistics are performed on the extracted pixel values ​​to calculate the mean of the pixels, which is used to reflect the overall brightness or color characteristics of the surrounding area;

[0043] Variance calculation: Calculate the variance of pixels in the surrounding unobstructed area to evaluate the distribution of image information and identify the magnitude of brightness or color changes;

[0044] Feature construction: Use the calculated mean and variance to infer the pixel values ​​of the occluded area to generate feature information of the occluded area.

[0045] Optionally, the behavior pattern recognition module includes a data receiving unit, a feature analysis unit, a pattern recognition unit and an abnormal behavior classification unit; wherein:

[0046] Data receiving unit: used to receive the biometric information from the spectral feature extraction module and the compensated image data from the occlusion detection compensation module;

[0047] Feature analysis unit: used to conduct in-depth analysis of the received biometric information and compensated image data, and extract the dynamic characteristics of the animal's posture, movement speed and range of activity;

[0048] Pattern recognition unit: uses machine learning algorithms to compare extracted features with known behavioral patterns to identify normal and abnormal behaviors of animals;

[0049] Abnormal behavior classification unit: Based on the pattern recognition results, the animal's behavior is classified as abnormal behavior such as anxiety, aggression or abnormal eating, and by setting a threshold, it is determined whether the recognition result meets the standard of abnormal behavior.

[0050] Optionally, the real-time warning module includes an abnormal behavior detection unit, a warning generation unit and a notification mechanism unit; wherein:

[0051] Abnormal behavior detection unit: used to receive abnormal behavior recognition results from the behavior pattern recognition module, and classify and record the recognized abnormal behaviors to form an abnormal behavior event log;

[0052] Warning generation unit: according to the recorded information in the abnormal behavior event log, a health warning threshold is set. When the recorded abnormal behavior exceeds the set threshold, a health warning message is automatically generated. The warning message includes the abnormal behavior type, occurrence time and recommended intervention measures;

[0053] Notification mechanism unit: The generated health warning information is immediately notified to the relevant veterinarian through the designated channel, and the designated channel includes SMS, email or mobile application push, so that the veterinarian can respond quickly.

[0054] Beneficial effects of the present invention:

[0055] The present invention significantly improves the accuracy of animal biological feature recognition through the organic combination of modules such as multi-view image acquisition, spectral feature extraction and occlusion detection compensation. The system can operate stably under different lighting conditions and effectively handle occlusion phenomena, thereby ensuring that the true state of the animal is captured and analyzed in real time. This precise biometric feature recognition can not only improve the ability to monitor the health status of animals, but also lay a solid foundation for subsequent behavioral pattern analysis.

[0056] The present invention uses the abnormal behavior detection function of the real-time behavior pattern recognition module, so that the system can quickly identify abnormal behaviors of animals such as anxiety, aggression or abnormal eating, and automatically generate health warnings. Combined with the intelligent notification mechanism, veterinarians can obtain warning information in time and intervene, which greatly improves the management efficiency of animal health. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0058] Figure 1 A schematic diagram of a visual inspection system according to an embodiment of the present invention;

[0059] Figure 2 Schematic diagram of an occlusion detection and compensation module according to an embodiment of the present invention. DETAILED DESCRIPTION

[0060] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0061] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0062] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0063] like Figure 1-Figure 2 As shown, a visual inspection system for animal husbandry and veterinary medicine includes a multi-view image acquisition module, a spectral feature extraction module, an occlusion detection and compensation module, a behavior pattern recognition module and a real-time warning module; wherein:

[0064] Multi-view image acquisition module: deploys multiple cameras to capture images of animals from different angles in real time to form an image dataset, providing basic data for subsequent analysis;

[0065] Spectral feature extraction module: used to process the image data set output by the multi-view image acquisition module, and use spectral feature extraction technology to analyze the reflectivity of the color and texture of animals at different wavelengths to extract the biological feature information of animals (such as fur color, texture, etc.) to support the rapid identification and tracking of individual animals;

[0066] Occlusion detection and compensation module: used to receive the image data set from the multi-view image acquisition module, identify the occlusion in the image by analyzing the motion information of the animal between different frames, and use the motion information to infer the features of the occluded part to make corresponding compensation, thereby improving the integrity of the image information and the recognition accuracy;

[0067] Behavior pattern recognition module: Based on the biometric information provided by the spectral feature extraction module and the image data compensated by the occlusion detection and compensation module, it analyzes the animal's behavior pattern and identifies the animal's abnormal behavior in real time, including anxiety, aggression or abnormal eating, to provide data support for veterinarians;

[0068] Real-time warning module: Based on the abnormal behavior recognition results of the behavior pattern recognition module, it automatically generates health warnings and notifies veterinarians to intervene.

[0069] The multi-view image acquisition module includes a camera deployment unit, an image capture unit and a data transmission unit; wherein:

[0070] Camera deployment unit: used to deploy high-definition cameras at predetermined locations in animal activity areas; specifically, cameras are installed at the four corners of the area to form a 360-degree monitoring view; in addition, two cameras with tilted angles are set in areas where animals frequently move to ensure that the behavior of animals at different heights and postures is captured. The installation height and angle of the camera are optimized according to the type and behavioral characteristics of the animal to ensure that each camera can cover its field of view;

[0071] Image capture unit: used to capture images of animals in real time and use automatic adjustment technology to automatically adjust the exposure value and white balance according to the ambient brightness under different lighting conditions. Specifically, a light sensor is used to monitor the ambient light in real time, and the camera settings are adjusted through image processing algorithms to maintain image clarity and color authenticity;

[0072] Data transmission unit: used to transmit the real-time captured image data to the subsequent processing module. The video encoding standard H.264 or H.265 is used in the transmission process to reduce the amount of data transmission. The specific transmission is achieved through optical fiber or Wi-Fi technology to ensure low latency and high stability. Through the organic coordination of the above units, the multi-view image acquisition module can realize all-round real-time monitoring of animals and form a complete image data set, providing a high-quality data foundation for subsequent behavior analysis, feature extraction and health assessment.

[0073] The spectral feature extraction module includes an image data receiving unit, a spectral analysis unit and a biometric information extraction unit; wherein:

[0074] Image data receiving unit: used to receive the image data set output from the multi-view image acquisition module, and format the received image data set to ensure the integrity and consistency of the data for subsequent analysis;

[0075] Spectral analysis unit: used to perform spectral decomposition on the received image data, analyze the reflectivity of animals at different wavelengths, and decompose the image into multiple spectral channels, so as to extract the color and texture characteristics of the animals and generate spectral feature maps. In this process, regularization technology is used to eliminate ambient light interference to ensure the accuracy of the extraction results;

[0076] Biometric information extraction unit: Based on the spectral feature map generated by the spectral analysis unit, the biometric information of the animal is analyzed and identified. The biometric information includes coat color, texture and skin condition. Then, the predetermined features are extracted by setting thresholds and pattern recognition algorithms, and stored in a preset biometric database for subsequent comparison and health status assessment.

[0077] The spectral analysis unit includes:

[0078] Spectral decomposition: First, the received image data is converted to the frequency domain through Fourier transform to identify the different frequency components in the image. This process can provide basic information for subsequent wavelength analysis;

[0079] Wavelength analysis: Set different wavelength ranges and process the image through filters to generate multiple spectral channels, each of which represents the reflectance characteristics of the animal at a predetermined wavelength, which is easy to separate and analyze;

[0080] Feature extraction: In the generated spectral channels, a clustering algorithm is used to extract the color and texture features of the animal. The mean and variance of the pixels in each channel are calculated to identify the feature areas, thereby extracting the animal's biological feature information, such as fur color and texture.

[0081] Spectral feature map generation: The color and texture features extracted at different wavelengths are combined to generate a spectral feature map, providing comprehensive biometric information for subsequent analysis;

[0082] Elimination of ambient light interference: During the entire process of spectral analysis unit operation, regularization technology is applied to eliminate ambient light interference to optimize the quality of image data and ensure the accuracy of extraction results. This measure helps to improve the reliability of final feature extraction. Through these steps, the spectral analysis unit can efficiently and accurately extract the biological characteristic information of animals, providing a solid data foundation for subsequent behavioral analysis and health management.

[0083] The operation steps of the above-mentioned spectrum analysis unit to generate a spectrum characteristic graph are as follows:

[0084] First, the received image data I(x, y) is spectrally decomposed, and the image is converted to the frequency domain using Fourier transform. The mathematical expression of Fourier transform is:

[0085] Among them, F(u, v) represents the frequency domain image, I(x, y) is the original image data, (x, y) is the image coordinate, (u, v) is the frequency variable, and j is the imaginary unit. The purpose of this step is to identify the different frequency components in the image through frequency domain analysis for subsequent wavelength analysis;

[0086] Next, by setting different wavelength ranges λ, the image is processed using filters to form multiple spectral channels S i (x, y), its expression is: S i (x, y) = I (x, y) · H (λ i ), where S i (x, y) is the i-th spectral channel image, H(λ i ) is for a specific wavelength λ i The filter function of this step is to separate the reflection characteristics of the image at different wavelengths in order to analyze the biological characteristics of the animal;

[0087] Then, multiple spectral channels are combined to extract the color and texture features of the animal through a clustering algorithm (such as K-means clustering). In each channel, the mean and variance of each pixel are calculated to identify the feature area. The formula is: Among them, μ is the mean of the feature points, σ 2 is the variance of the feature points, N is the number of feature points, (x j ,y j ) is the coordinate of the jth feature point. Through this step, the biological characteristics of the animal, such as fur color and texture information, can be effectively extracted;

[0088] Finally, the color and texture features at different wavelengths are combined to generate the spectral feature map G(x, y), which is expressed as: Among them, G(x, y) is the generated spectral feature map, w i is the weight coefficient of the i-th spectral channel, M is the number of spectral channels. The purpose of this step is to integrate the information at different wavelengths to generate a comprehensive biological feature map and provide a data basis for subsequent analysis;

[0089] In the whole process, regularization techniques (such as L2 regularization) are used to eliminate ambient light interference, and the optimization objective function is: Among them, I is the original image, I clean is the image after removing ambient light, λ is the regularization parameter, which represents the weight of denoising. The purpose of this step is to improve the accuracy of feature extraction and ensure that the final result is not affected by the external environment.

[0090] The occlusion detection and compensation module includes a motion information extraction unit, an occlusion detection unit, a feature inference unit, and a compensation generation unit; wherein:

[0091] Motion information extraction unit: used to receive the image data set from the multi-view image acquisition module, and use the Lucas-Kanade algorithm to calculate the motion vector of the animal between different frames, identify the motion direction and speed of each pixel, and thus obtain the movement trajectory and dynamic behavior of the animal;

[0092] Occlusion detection unit: By analyzing motion information, comparing motion patterns between different frames, establishing an optical flow field, and identifying the mutation area of ​​the motion vector, the mutation area corresponds to the part occluded by other objects, so as to accurately detect the occlusion in the image;

[0093] Feature inference unit: used to infer the features of the occluded part using motion information after identifying the occlusion situation;

[0094] Compensation generation unit: It is used to combine the inferred features with the actual image. The generated compensated image can be seamlessly embedded into the original image, thereby restoring the complete appearance of the animal and reducing the information loss caused by occlusion. Through the collaborative work of the above units, the occlusion detection compensation module can efficiently and accurately identify and process occlusion situations, providing more complete visual data for subsequent behavioral analysis and health monitoring.

[0095] The motion information extraction unit includes:

[0096] Image frame acquisition: First, continuous image frames are acquired from the multi-view image acquisition module to analyze the state of the animal at different time points;

[0097] Optical flow calculation: Lucas-Kanade algorithm is used to calculate the motion vector of each pixel by analyzing the image brightness changes between adjacent frames, and identify its displacement in the x and y directions;

[0098] Least squares solution: To accurately obtain the motion vector, a linear equation containing multiple neighborhood pixels is constructed and solved using the least squares method to obtain the motion direction and speed of each pixel, thereby forming complete motion vector data;

[0099] Movement trajectory acquisition: Using the calculated motion vector data, the animal's movement trajectory is formed through a tracking algorithm to reflect the animal's dynamic behavior in real time; through these steps, the motion information extraction unit can accurately identify the animal's movement pattern, providing important data support for subsequent occlusion detection and behavior analysis.

[0100] The calculation steps to identify the direction and speed of motion of each pixel are as follows:

[0101] First, continuous image frames are acquired from the multi-view image acquisition module, which are set as I(x, y, t) and I(x, y, t+Δt), where I represents image intensity, x and y represent image coordinates, t represents time point, and Δt represents time interval;

[0102] Next, the Lucas-Kanade algorithm is used to construct the following optical flow equation by assuming the smoothness of the optical flow and the invariance of the image brightness: x u+I y v+I t =0, where I x , I y and I t Respectively represent the gradient of the image in the x, y and time directions, u and v represent the optical flow components in the x and y directions respectively;

[0103] Then: To solve the motion vector (u, v) of each pixel, a linear equation system containing multiple neighborhood pixels is constructed and solved using the least squares method. The equation system is expressed as: A·uv=b, where A is the image gradient matrix and b is the time gradient vector;

[0104] Next: By solving the above equations, we can obtain the movement direction and speed of each pixel, i.e. the motion vector (u i , v i );

[0105] Finally: Based on the calculated motion vector, the moving trajectory of the animal between image frames is tracked by integration to form the dynamic behavior of the animal.

[0106] The occlusion detection unit specifically includes:

[0107] Optical flow field construction: First, the motion vector data provided by the motion information extraction unit is used to construct the optical flow field, which is used to represent the motion direction and speed of each pixel between different frames;

[0108] Motion pattern comparison: compare the optical flow fields between adjacent frames, identify changes in motion patterns, and determine areas where motion vectors change suddenly, especially focusing on areas that are inconsistent with surrounding motion vectors;

[0109] Identification of mutation areas: Using the set threshold, the motion vectors are filtered to identify areas where the motion vectors change significantly. These areas indicate the occurrence of occlusion. Through the above steps, the occlusion detection unit can accurately identify occlusion during animal movement, providing an important basis for subsequent behavioral analysis.

[0110] The features of the occluded area constructed in the feature inference unit specifically include:

[0111] Unoccluded area identification: First identify the unoccluded pixel area to ensure that the extracted data is not affected by occlusion;

[0112] Pixel value extraction: Extract pixel values ​​from unobstructed areas to provide basic data for subsequent analysis;

[0113] Mean calculation: Statistics are performed on the extracted pixel values ​​to calculate the mean of the pixels, which is used to reflect the overall brightness or color characteristics of the surrounding area;

[0114] Variance calculation: Calculate the variance of pixels in the surrounding unobstructed area to evaluate the distribution of image information and identify the magnitude of brightness or color changes;

[0115] Feature construction: Use the calculated mean and variance to infer the pixel value of the occluded area. The inferred value is based on the mean of the surrounding unobstructed area and takes into account the changes reflected by the variance to generate feature information of the occluded area. Through the above steps, the feature inference unit can effectively infer the features of the occluded area.

[0116] The specific steps to construct the features of the occluded area are as follows:

[0117] First, the unobstructed pixel area is identified through the information provided by the occlusion detection unit to ensure that the extracted data is not affected by occlusion; and the pixel values ​​of the unobstructed area are extracted. These pixel values ​​are used for subsequent statistical analysis to ensure that only valid information is included in the data analysis;

[0118] Then, the extracted pixel values ​​are counted and the pixel mean of the surrounding unobstructed area is calculated. The formula is: Among them, I i Represents the value of each pixel in the unoccluded area, and N is the total number of unoccluded pixels;

[0119] Next, the pixel variance of the surrounding unoccluded area is calculated to evaluate the distribution of image information. The formula is:

[0120] Finally, the calculated mean and variance are used to infer the pixel values ​​of the occluded area. It is approximately the mean of the pixel values ​​in the surrounding unoccluded area, plus a deviation term related to the variance, expressed as: Among them, k is an empirical coefficient that reflects the similarity between the features of the occluded area and the surrounding area.

[0121] The behavior pattern recognition module includes a data receiving unit, a feature analysis unit, a pattern recognition unit and an abnormal behavior classification unit; wherein:

[0122] Data receiving unit: used to receive the biometric information from the spectral feature extraction module and the compensated image data from the occlusion detection and compensation module to ensure the integrity and accuracy of the input data;

[0123] Feature analysis unit: used to perform in-depth analysis on the received biometric information and compensated image data, and extract the dynamic features of the animal's posture, movement speed and range of activity; for example, the animal's movement speed v is calculated by the formula Where d is the displacement and t is the time;

[0124] Pattern recognition unit: using machine learning algorithms (such as support vector machines or decision trees) to compare the extracted features with known behavior patterns, and identifying the normal and abnormal behaviors of animals through the trained model; for example, based on the support vector machine algorithm, the extracted features are classified by constructing a hyperplane, and the specific steps are as follows:

[0125] 1. Map the data into a high-dimensional space to improve separability;

[0126] 2. Calculate the support vector to define the classification boundary;

[0127] 3. Finally, the normal and abnormal behaviors of animals are identified and the confidence of the classification results is calculated to determine its accuracy.

[0128] Abnormal behavior classification unit: Based on the pattern recognition results, the animal's behavior is classified as abnormal behavior such as anxiety, aggression or abnormal eating, and a threshold is set to determine whether the recognition result meets the standard of abnormal behavior; the specific judgment steps are as follows:

[0129] Feature threshold setting: First, set the feature threshold for each behavior category. For example, the identification criteria for anxious behavior may include activity range, heart rate, and behavior frequency.

[0130] Activity range calculation: Calculate the animal's activity range R using the formula: Among them, R is the range of activity, x i For the location information of each observation point, is the average position of all observation points;

[0131] Abnormal behavior determination: Compare the calculated activity range R with the set threshold T. If R>T, it is determined to be anxiety behavior. For other abnormal behaviors, set corresponding judgment conditions. For example, the identification of aggressive behavior can be based on action frequency and speed.

[0132] Output feedback: Based on the classification results, feedback information is generated to prompt the veterinarian to make necessary observations or interventions. For example, the output information may be "the animal exhibits anxious behavior, and further observation is recommended."

[0133] Through the above steps, the behavioral pattern recognition module can analyze the animal's behavioral patterns in real time, quickly identify abnormal behaviors, and provide timely decision support for veterinarians. This process combines biometric information and compensated image data to ensure the accuracy and reliability of recognition.

[0134] The real-time warning module includes an abnormal behavior detection unit, a warning generation unit, and a notification mechanism unit; wherein:

[0135] Abnormal behavior detection unit: used to receive abnormal behavior recognition results from the behavior pattern recognition module, and classify and record the recognized abnormal behaviors to form an abnormal behavior event log;

[0136] Warning generation unit: according to the recorded information in the abnormal behavior event log, set the health warning threshold. When the recorded abnormal behavior exceeds the set threshold, the health warning information is automatically generated. The warning information includes the abnormal behavior type, occurrence time and recommended intervention measures.

[0137] Notification mechanism unit: The generated health warning information will be notified to the relevant veterinarians in real time through designated channels, including SMS, email or mobile application push, so that the veterinarians can respond quickly; through the above steps, the real-time warning module can effectively monitor the abnormal behavior of animals and notify veterinarians in time, ensuring real-time attention and intervention on the health status of animals, and improving the efficiency and safety of animal management.

[0138] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0139] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A visual inspection system for animal husbandry and veterinary medicine, characterized in that: It includes a multi-view image acquisition module, a spectral feature extraction module, an occlusion detection and compensation module, a behavior pattern recognition module and a real-time warning module; among which: Multi-view image acquisition module: deploys multiple cameras to capture images of animals from different angles in real time to form an image dataset; Spectral feature extraction module: used to process the image data set output by the multi-view image acquisition module, and use spectral feature extraction technology to analyze the reflectivity of the color and texture of the animal at different wavelengths to extract the biological feature information of the animal; Occlusion detection and compensation module: used to receive the image data set from the multi-view image acquisition module, identify the occlusion in the image by analyzing the motion information of the animal between different frames, and use the motion information to infer the features of the occluded part to make corresponding compensation; Behavior pattern recognition module: Based on the biometric information provided by the spectral feature extraction module and the image data compensated by the occlusion detection and compensation module, it analyzes the animal's behavior pattern and identifies the animal's abnormal behavior in real time, including anxiety, aggression or abnormal eating; Real-time warning module: Based on the abnormal behavior recognition results of the behavior pattern recognition module, it automatically generates health warnings and notifies veterinarians to intervene.

2. A visual inspection system for animal husbandry and veterinary medicine according to claim 1, characterized in that: The multi-view image acquisition module includes a camera deployment unit, an image capture unit and a data transmission unit; wherein: Camera deployment unit: used to deploy high-definition cameras at predetermined locations in animal activity areas; specifically, cameras are installed at the four corners of the area to form a 360-degree monitoring view; in addition, two cameras with tilted angles are set in areas where animals frequently move; Image capture unit: used to capture images of animals in real time, and uses automatic adjustment technology to automatically adjust the exposure value and white balance according to the ambient brightness under different lighting conditions; Data transmission unit: used to transmit the real-time captured image data to the subsequent processing module. The video encoding standard H.264 or H.265 is used in the transmission process to reduce the data transmission volume. The specific transmission is achieved through optical fiber or Wi-Fi technology.

3. A visual inspection system for animal husbandry and veterinary medicine according to claim 1, characterized in that: The spectral feature extraction module includes an image data receiving unit, a spectral analysis unit and a biometric information extraction unit; wherein: Image data receiving unit: used to receive the image data set output from the multi-view image acquisition module, and format the received image data set to ensure the integrity and consistency of the data; Spectral analysis unit: used to perform spectral decomposition on the received image data, analyze the reflectivity of the animal at different wavelengths, and decompose the image into multiple spectral channels, so as to extract the color and texture characteristics of the animal and generate a spectral feature map; Biometric information extraction unit: Based on the spectral feature graph generated by the spectral analysis unit, analyze and identify the animal's biometric information, the biometric information including fur color, texture and skin condition.

4. A visual inspection system for animal husbandry and veterinary medicine according to claim 1, characterized in that: The spectrum analysis unit comprises: Spectral decomposition: The received image data is first converted to the frequency domain through Fourier transform to identify the different frequency components in the image; Wavelength analysis: Set different wavelength ranges and process the image through filters to generate multiple spectral channels, each of which represents the reflectance characteristics of the animal at a predetermined wavelength; Feature extraction: In the generated spectral channels, a clustering algorithm is used to extract the color and texture features of the animal, and the mean and variance of the pixels in each channel are calculated to identify the characteristic areas, thereby extracting the biological characteristic information of the animal; Spectral feature map generation: The color and texture features extracted at different wavelengths are integrated to generate a spectral feature map; Ambient light interference elimination: During the entire process of spectral analysis unit operation, regularization techniques are applied to eliminate ambient light interference to optimize the quality of image data.

5. The animal husbandry and veterinary visual inspection system according to claim 1, characterized in that: The occlusion detection and compensation module includes a motion information extraction unit, an occlusion detection unit, a feature inference unit and a compensation generation unit; wherein: Motion information extraction unit: used to receive the image data set from the multi-view image acquisition module, and use the Lucas-Kanade algorithm to calculate the motion vector of the animal between different frames, identify the motion direction and speed of each pixel, and thus obtain the movement trajectory and dynamic behavior of the animal; Occlusion detection unit: By analyzing motion information, comparing motion patterns between different frames, establishing an optical flow field, and identifying the mutation area of ​​the motion vector, the mutation area corresponds to the part occluded by other objects, so as to accurately detect the occlusion in the image; Feature inference unit: used to infer the features of the occluded part using motion information after identifying the occlusion situation; Compensation generation unit: used to combine the inferred features with the actual image. The generated compensated image can be seamlessly embedded into the original image, thereby restoring the complete appearance of the animal and reducing the information loss caused by occlusion.

6. A visual inspection system for animal husbandry and veterinary medicine according to claim 5, characterized in that: The motion information extraction unit comprises: Image frame acquisition: first obtain continuous image frames from the multi-view image acquisition module; Optical flow calculation: Lucas-Kanade algorithm is used to calculate the motion vector of each pixel by analyzing the image brightness changes between adjacent frames, and identify its displacement in the x and y directions; Least squares solution: Construct a linear equation containing multiple neighborhood pixels and use the least squares method to solve it, so as to obtain the movement direction and speed of each pixel, and then form complete motion vector data; Movement trajectory acquisition: Using the calculated motion vector data, the animal's movement trajectory is formed through a tracking algorithm to reflect the animal's dynamic behavior in real time.

7. A visual inspection system for animal husbandry and veterinary medicine according to claim 6, characterized in that: The occlusion detection unit specifically includes: Optical flow field construction: First, the motion vector data provided by the motion information extraction unit is used to construct the optical flow field, which is used to represent the motion direction and speed of each pixel between different frames; Motion pattern comparison: compare the optical flow fields between adjacent frames, identify changes in motion patterns, and determine areas where motion vectors change suddenly, especially focusing on areas that are inconsistent with surrounding motion vectors; Mutation area identification: Use the set threshold to filter the motion vector and identify the area where the motion vector changes significantly.

8. A visual inspection system for animal husbandry and veterinary medicine according to claim 7, characterized in that: The feature inference unit specifically includes: Unoccluded area recognition: first identify the unoccluded pixel area; Pixel value extraction: Extract pixel values ​​from unobstructed areas to provide basic data for subsequent analysis; Mean calculation: Statistics are performed on the extracted pixel values ​​to calculate the mean of the pixels, which is used to reflect the overall brightness or color characteristics of the surrounding area; Variance calculation: Calculate the variance of pixels in the surrounding unobstructed area to evaluate the distribution of image information and identify the magnitude of brightness or color changes; Feature construction: Use the calculated mean and variance to infer the pixel values ​​of the occluded area to generate feature information of the occluded area.

9. The animal husbandry and veterinary visual inspection system according to claim 1, characterized in that: The behavior pattern recognition module includes a data receiving unit, a feature analysis unit, a pattern recognition unit and an abnormal behavior classification unit; wherein: Data receiving unit: used to receive the biometric information from the spectral feature extraction module and the compensated image data from the occlusion detection compensation module; Feature analysis unit: used to conduct in-depth analysis of the received biometric information and compensated image data, and extract the dynamic characteristics of the animal's posture, movement speed and range of activity; Pattern recognition unit: uses machine learning algorithms to compare extracted features with known behavioral patterns to identify normal and abnormal behaviors of animals; Abnormal behavior classification unit: Based on the pattern recognition results, the animal's behavior is classified as abnormal behavior such as anxiety, aggression or abnormal eating, and by setting a threshold, it is determined whether the recognition result meets the standard of abnormal behavior.

10. The animal husbandry and veterinary visual inspection system according to claim 1, characterized in that: The real-time warning module includes an abnormal behavior detection unit, a warning generation unit and a notification mechanism unit; wherein: Abnormal behavior detection unit: used to receive abnormal behavior recognition results from the behavior pattern recognition module, and classify and record the recognized abnormal behaviors to form an abnormal behavior event log; Warning generation unit: according to the recorded information in the abnormal behavior event log, a health warning threshold is set. When the recorded abnormal behavior exceeds the set threshold, a health warning message is automatically generated. The warning message includes the abnormal behavior type, occurrence time and recommended intervention measures; Notification mechanism unit: The generated health warning information is immediately notified to the relevant veterinarian through the designated channel, and the designated channel includes SMS, email or mobile application push, so that the veterinarian can respond quickly.

Images