Beef cattle oestrus state monitoring method based on dynamic gait analysis

Abstract

The invention relates to the technical field of behavior feature recognition, in particular to a beef cattle oestrus state monitoring method based on dynamic gait analysis. According to the method, multi-source data acquisition, such as dynamic gait video and body temperature data, video preprocessing, such as frame extraction, denoising, abnormal frame recognition and restoration, target tracking, dynamic gait characteristic parameter extraction, such as step length, step frequency, step speed, joint angle change and body gravity center offset, and life data acquisition, such as diet, body weight change and drinking water. A large oestrus monitoring and recognition model integrating attitude analysis, behavior analysis, living state analysis and an integration module is constructed, and after training and iterative optimization, accurate judgment of oestrus, non-oestrus and wig oestrus states of beef cattle is achieved in combination with ear tag identity recognition. By monitoring the dynamic gaits and behavior modes of the beef cattle and combining daily life data of the beef cattle, the problem of misjudgment caused by wig oestrus is greatly avoided, and the monitoring accuracy is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of behavior feature recognition, and in particular to a beef cow estrus state monitoring method based on dynamic gait analysis. BACKGROUND

[0002] In large-scale beef cattle breeding, accurately identifying the estrus state of beef cattle is a key link to improve reproductive efficiency; traditional monitoring methods rely on manual observation (such as observing beef cattle's mounting behavior and vulva changes), which has the following defects: 1. Low efficiency: a single feeder can only monitor 50-80 beef cattle per day, with a false negative rate of up to 25%-30% (referring to data from China Animal Husbandry Journal, Vol. 59, 2023); 2. Strong subjectivity: the misjudgment rate of "false estrus" (such as abnormal behavior caused by stress) is over 15%; 3. Single data: without combining dynamic physiological indicators such as gait and body temperature, the behavior change pattern of beef cattle during estrus cannot be quantified.

[0003] In the prior art, some solutions use a single sensor (such as a body temperature sensor or an activity meter) for monitoring, but there are limitations: only through body temperature fluctuations (estrus period body temperature rises by 0.3-0.5℃) for identification, the accuracy is less than 80%; only through activity level, it is easily disturbed by feed replacement, environmental changes, etc., with a misjudgment rate of over 20%.

[0004] Therefore, the present application proposes a multi-dimensional data fusion monitoring method, which quantifies dynamic gait features, repairs abnormal data, and iteratively optimizes the model to solve the problems of "low accuracy and weak anti-interference ability" in the prior art, achieving an estrus state recognition accuracy of ≥98% and a life data interference rate of ≤5%. SUMMARY

[0005] The purpose of the present application is to solve the problems in the prior art and to propose a beef cow estrus state monitoring method based on dynamic gait analysis.

[0006] In order to achieve the above purpose, the present application adopts the following technical scheme: a beef cow estrus state monitoring method based on dynamic gait analysis includes the following steps: S1: Collecting beef cattle dynamic gait video and detecting beef cattle body temperature in real time to obtain body temperature data; S2: Preprocessing the collected dynamic gait video, including video frame extraction, image denoising, target detection and tracking, to obtain a continuous gait image sequence of a single beef cattle, and marking real-time body temperature data; analyzing the dynamic gait video through an analysis module to obtain beef cattle behavior data; S3: extracting dynamic gait feature parameters of the beef cattle from the continuous gait image sequence, the dynamic gait feature parameters including step length, step frequency, step speed, joint angle change, and body center of gravity offset; S4: obtaining life data by recording daily work of beef cattle diet data and body weight change; S5: inputting the extracted dynamic gait feature parameters, behavior data, body temperature data and life data into an estrus monitoring and recognition large model for training; S6: obtaining an identification result of the estrus state of the corresponding beef cattle by inputting the dynamic gait feature parameters, behavior data, body temperature data and life data obtained by ear tag recognition of the cattle identity information into the estrus monitoring and recognition large model for analysis, the identification result including estrus, non-estrus and false estrus.

[0007] Preferably, the life data in S4 further includes beef cattle water drinking data.

[0008] Preferably, the specific process of extracting the joint angle change in S3 is: identifying key joint points of the beef cattle through a skeleton key point detection algorithm, calculating the angle change between the key joint points in adjacent frames, and obtaining joint angle change data.

[0009] Preferably, the dynamic gait video is a multi-angle video obtained by setting multiple fixed points of high-definition cameras and multiple continuously movable positions of mobile cameras in a cowshed.

[0010] Preferably, the estrus monitoring and recognition large model includes a posture analysis module, a behavior analysis module, a life state analysis module and an integration module; wherein the posture analysis module receives the dynamic gait feature parameters of the beef cattle, analyzes the cattle posture and joint change information; the behavior analysis module receives the behavior data, analyzes the cattle behavior mode information; the life state analysis module receives the life data, analyzes the beef cattle state according to the cattle diet and water drinking and body weight change data; and the integration module integrates and analyzes the analysis results of the posture analysis module, the behavior analysis module and the life state analysis module to avoid the influence of life data on the estrus of beef cattle.

[0011] Preferably, the preprocessing of the dynamic gait video in S2 further includes an abnormal frame identification and repair step: constructing an abnormal frame detection model through a convolutional neural network (CNN), determining the blur degree and occlusion rate of the extracted video frames, and marking as an abnormal frame when the blur degree > 30% or the occlusion rate > 40%; using a step feature interpolation algorithm of adjacent 3 frames to repair the abnormal frame, generating a supplementary frame conforming to the continuous gait logic, replacing the original abnormal frame and then performing target tracking.

[0012] Preferably, the estrus monitoring recognition large model further comprises an iterative optimization module: the module collects not less than 50 groups of complete data of verified estrus / non-estrus beef cattle per month, including dynamic gait feature parameters, behavior data, body temperature data and life data, and fine tunes the weight parameters of the posture analysis module and the behavior analysis module through the small batch gradient descent method; meanwhile, false estrus misjudgment as estrus is recorded as a model misjudgment case, the weight distribution rules of the attribution analysis correction integration module are modified, so that the interference rate of life data on estrus determination is reduced to below 5%.

[0013] The beneficial effects of the present application are as follows: 1. The present application has high precision through multi-modal fusion: it breaks through the limitations of single sensor monitoring, significantly improves the accuracy of estrus recognition through cross-validation of four-dimensional data of gait, behavior, body temperature and life, and effectively distinguishes between "true estrus" and "false estrus"; 2. The present application has small stress through non-contact monitoring: the computer vision-based analysis method does not require physical contact with the cattle, reduces the stress reaction of animals, and is more in line with animal welfare requirements; 3. The present application is automated and intelligent: it realizes 24-hour all-weather unattended monitoring, greatly reduces labor costs, and makes the system have the ability of continuous evolution through the iterative optimization module. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 The method flowchart of the present application. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical scheme and advantages of the present application clearer and more apparent, the present application will be further described in detail below with reference to the drawings. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0016] Please refer to Figure 1 The present application provides a technical scheme: a beef cattle estrus state monitoring method based on dynamic gait analysis includes the following steps: S1: collecting beef cattle dynamic gait video and detecting beef cattle body temperature in real time to obtain body temperature data; 3-5 fixed point 4K high-definition cameras (frame rate 25fps, resolution 3840x2160) are arranged in the cowshed, and 2 mobile track cameras (moving speed 0.5m / s, covering the blind area of the cowshed) are configured to collect multi-angle videos of beef cattle walking, standing and moving; Body temperature data acquisition: an ear tag type wireless temperature sensor (measurement range 36-42℃, accuracy ±0.1℃, sampling frequency 1 / time 5min) is used to obtain beef cattle body temperature data T(t) in real time, where t is the collection time stamp (unit: min); S2: Preprocess the acquired dynamic gait videos, including video frame extraction, image denoising, target detection and tracking, to obtain a continuous gait image sequence of a single beef cattle, and label real-time body temperature data; analyze the dynamic gait videos through the analysis module to obtain beef cattle behavior data; Video frame extraction requires "timestamp alignment + keyframe filtering" to ensure the temporal continuity and validity of gait data. The formula is as follows: Frame extraction time interval calculation To balance data volume and gait integrity, the extraction interval needs to be dynamically adjusted according to the moving speed of the beef cattle. The formula is as follows: Frame extraction time interval (unit: s, default ≤0.05s, i.e. extraction frequency of 20fps or higher). F: Camera's original frame rate (30fps, fixed value); Minimum stride length for beef cattle (S_min = 0.8m for adult beef cattle, based on historical data statistics); Safety factor (1.2, to avoid gait frame loss due to step size fluctuations); Example: When F=30fps and S_min=0.8m, =1 / (30×0.8)×1.2=0.05s, that is, 1 frame is extracted every 0.05s, ensuring that each step contains at least 3 keyframes; Timestamp Alignment Formula Video captured by multiple cameras (fixed + moving) needs to have a unified time base and be corrected using UTC timestamps. The formula is as follows: : Uniform timestamp after synchronization (unit: s, accurate to milliseconds); : Camera's original timestamp (there may be clock discrepancies between devices); Time offset (calibrated via NTP network time protocol) ≤10ms, ensuring multi-view frame time difference <0.01s); Image denoising (Gaussian noise and salt-and-pepper noise suppression) To address image noise caused by variations in cowshed lighting (such as differences in light levels between day and night) and interference from camera sensors, a combination of Gaussian filtering and median filtering is used for noise reduction. The formula is as follows: Gaussian filtering denoising formula (suppressing Gaussian noise) The extracted video frames are then subjected to neighborhood-weighted smoothing, using the following formula: : the gray value of the (x, y) pixel after denoising; : the gray value (0-255) of the (x, y) pixel of the original noise frame; k: filter kernel size parameter (k=1, i.e., 3x3 filter kernel, balancing denoising effect and edge preservation); : Gaussian function standard deviation (adaptively adjusted according to noise intensity, =1.2-1.8, default in the cowshed scene =1.5, the more serious the noise , the larger); : Gaussian weight coefficient (the closer the neighborhood pixel is to the center, the larger the weight, to avoid edge blurring); Physical meaning: Through the Gaussian weighted average of neighborhood pixels, random distributed Gaussian noise (such as sensor thermal noise) is suppressed, and the noise variance after denoising is reduced to less than 15% of the original variance.

[0017] Median filter denoising formula (suppressing salt and pepper noise) Further processing of the salt and pepper noise (such as black and white noise points caused by dust obstruction and light sudden change) remaining after Gaussian filtering, the formula is: : the final gray value of the (x, y) pixel after denoising; : median calculation function (sort the gray values of the 9 pixels in the 3x3 neighborhood, and take the middle value as the current pixel value); : index range of 3x3 neighborhood (covering the current pixel and the surrounding 8 pixels); Advantages: Compared with mean filter, median filter can remove salt and pepper noise (best effect when noise ratio ≤20%) while preserving the edge information of key gait features such as beef cattle outline and hoof, with edge clarity preservation rate ≥90%.

[0018] Target detection and tracking (beef cattle individual recognition and trajectory association) Based on YOLOv8 target detection algorithm, beef cattle positioning is realized, combined with Kalman filter for multi-frame trajectory tracking to ensure the accuracy of the continuous gait image sequence of a single cow, the formula is as follows: YOLOv8 target detection confidence calculation (beef cattle region determination) Beef cattle target detection is performed on the denoised image, and the confidence of the target region is output, the formula is: C: Final confidence level of the target area (C≥0.7 is considered a valid beef cattle area to avoid false detection of background objects such as feed troughs and fences); : Confidence score of the "beef cattle" category output by the YOLOv8 model (initial confidence score, range [0,1]); M: Number of matches between the target candidate box and the ground truth box (M=3, taking the 3 candidate boxes with the highest matching scores); IOU(m): Intersection over Union (IoU) ratio between the m-th candidate box and the ground truth box (measuring the accuracy of target localization). , The overlapping area, (To merge areas, an IOU(m) ≥ 0.5 is considered a valid match). Example: If =0.85, the IOU of the three candidate boxes are 0.6, 0.55 and 0.5 respectively, then C=0.85×0.6=0.51 (take the maximum product). Since C<0.7, the detection needs to be repeated to ensure the detection accuracy is ≥98%. Kalman filter tracking formula (trajectory prediction and association) Trajectory tracking of cattle targets in consecutive frames achieves a closed loop of "detection-prediction-update," and the formula consists of two steps: (1) State prediction (predicting the target position in the current frame based on the trajectory of the previous frame): (2) State update (correcting the predicted trajectory based on the detection results of the current frame): Note: The predicted value (state vector) of the target state in the current frame (k frames). x, y are the coordinates of the target center, and w, h are the width and height of the target bounding box. (The speed of movement in the x and y directions). State transition matrix ( (Δt is the frame interval, describing the relationship between state changes over time). The optimal estimate of the target state in the previous frame (frame k-1); : Control input matrix ( =0, meaning there is no active control input in the cattle shed scenario (e.g., no external force pushes the cattle to move). : Control input vector (empty if no control input) =0, omitted if no control input : Covariance matrix of current frame state prediction (measures prediction uncertainty, initialized as diagonal matrix with diagonal elements [5, 5, 2, 2, 1, 1] in pixel 2 , pixel 2 , pixel 2 , pixel 2 , (pixel / s) 2 , (pixel / s) 2 ) : Transpose of state transition matrix : Process noise covariance matrix =0.01×I, I is a 6-order identity matrix, controls prediction error, avoids trajectory drift : Kalman gain (balances reliability of predicted value and detected value, greater, more dependent on detected value smaller, more dependent on predicted value : Observation matrix maps 6-dimensional state vector to 4-dimensional observation vector, observation value only contains target position and size : Transpose of observation matrix : Observation value of current frame target (i.e. target box coordinates detected by YOLOv8 ) : Observation noise covariance matrix =0.1×I, I is a 4-order identity matrix, reflects detection error, set according to YOLOv8 detection accuracy : Inverse of observation residual covariance matrix : Optimal estimate value of current frame target state (position and velocity of final tracking trajectory) : Identity matrix (dimension consistent with state vector, 6-order) : Covariance matrix of optimal estimate of current frame state (uncertainty reduced after update, ensures trajectory continuity) ​Tracking effect: In the multi-cow scenario (20 beef cattle), the trajectory association accuracy is ≥95%, there is no identity confusion (combined with RFID ear tag ID for further calibration), and the tracking interruption rate is <1% (short-term interruption due to obstruction can be repaired by trajectory interpolation).

[0019] Behavior data extraction: Identify beef cattle behavior through a behavior classification model (based on CNN-LSTM architecture) and output behavior data , where b takes the value of "standing (0), walking (1), lying (2), and active (3)", and the time length proportion of each behavior is calculated (such as walking behavior proportion ); S3: Extract dynamic gait feature parameters of beef cattle from the continuous gait image sequence, including step length, step frequency, step speed, joint angle change, and body center of gravity offset; Continuous gait image sequence 5 core parameters are extracted: Step length (L): Through cow hoof key point detection (using an improved version of OpenPose algorithm to identify cow hoof tip and hoof heel 2 key points), calculate the Euclidean distance between adjacent two-step hoof tips, formula as follows: Where ( , ) is the i-th step hoof tip coordinate, ( , ) is the i+1 step hoof tip coordinate; take the average of 10 consecutive steps as the step length of a single cow ; Step frequency (f): Calculate the number of steps per unit time, formula as follows: Where N is the number of steps in T time (T takes 5 min), unit is steps / min.

[0020] Step speed (v): Combine step length and step frequency, formula as follows: Unit is m / min (step length unit conversion to m).

[0021] Joint angle change ( ): Identify 3 key joint points of beef cattle hip joint, knee joint, and elbow joint, construct joint vectors (such as hip joint vector , knee joint vector ), calculate joint angle through vector dot product, formula as follows: The angle difference of adjacent frames is the joint angle change , and the average value of 20 consecutive frames is taken as .

[0022] Body center of gravity offset (D): taking the beef cattle torso center as the reference point , the Euclidean distance between the torso center of each frame and the reference point is calculated, and the formula is as follows: Take the maximum offset of 15 consecutive frames as the center of gravity offset feature parameter.

[0023] S4: Record the beef cattle diet data and body weight changes through daily work to obtain life data; Record the beef cattle life data L=\{E,W,H\} through the farm management system: Diet data (E): record the average daily intake of single cattle (unit: kg) by using automatic feeding system , and the formula is as follows: Where is the i-th intake, and n is the average daily intake frequency (usually 3 times); Body weight change (W): measure the body weight of beef cattle through the body weight scale every week , and the body weight change formula is as follows: The unit is kg; Water consumption data (H): record the average daily water consumption of single cattle (unit: L) by using intelligent water dispenser , and the calculation method is the same as the intake .

[0024] S5: input the extracted dynamic gait feature parameters, behavior data, body temperature data and life data into the estrus monitoring and recognition large model for training; Construct a multi-module fusion estrus monitoring and recognition large model, the model input is "dynamic gait feature parameters , behavior data B, body temperature data T(t), life data L", and the output is estrus state (estrus=1, non-estrus=0, false estrus=2), and the specific module design is as follows: Posture analysis module: adopt full connection neural network (input layer 5 neurons, hidden layer 2 layers each 32 neurons, output layer 1 neuron), output posture score , and the formula is as follows: Where The weight matrix is optimized by training, The bias term, The Sigmoid activation function , >0.8 indicates that the posture meets the estrus characteristics; Behavior analysis module: use LSTM network to analyze behavior data B, output behavior score , The formula is as follows: Among them is the time sequence feature output of the behavior sequence, , is the training parameter ( >0.75 indicates that the behavior meets the estrus characteristics; Life state analysis module: calculate the life data deviation D_L (measure the interference of life data on estrus judgment), the formula is as follows: Among them is the average value of life data of beef cattle in non-estrus period (calculated by 1000 head of beef cattle historical data), <0.1 indicates that there is no significant interference of life data.

[0025] Integration module: integrate the outputs of the three modules to calculate the final estrus judgment score , The formula is as follows: Among them =0.4、 =0.4、 =0.2 (initial weight, adjusted by iteration optimization); Judgment rules: >0.7: Estrus; 0.3≤ ≤0.7: False estrus; <0.3: Non-estrus.

[0026] Model training process: 8000 groups of beef cattle data (including 4000 groups of estrus data, 3500 groups of non-estrus data, and 500 groups of false estrus data) are used to optimize parameters with cross entropy loss function , The formula is as follows: Among them is the true label (estrus=1, non-estrus=0, false estrus=2, one-hot encoding is used for multi-classification), the number of training iterations is 100 rounds, and the learning rate is 0.001. The model accuracy rate reaches 97.5% after training; S6: Identify the identity information of the cow through the ear mark, input the acquired dynamic gait feature parameters, behavior data, body temperature data and life data into the estrus monitoring and recognition large model for analysis, and obtain the recognition result corresponding to the estrus state of the beef cow, the recognition result including estrus, non-estrus and false estrus.

[0027] Identity recognition: reading the identity information ID of the beef cow through the ear tag RFID module, and associating the historical data (such as non-estrus period gait mean value, life data mean value) of the cow; Real-time analysis: inputting the real-time collected G, B, T(t), L into the trained model, outputting , obtaining the estrus state result according to the determination rule, and pushing to the breeder terminal through the breeding farm management platform.

[0028] As an embodiment of the present application, the life data in S4 further includes beef drinking water data.

[0029] As an embodiment of the present application, the specific process of extracting the joint angle change in S3 is: identifying the key joint points of the beef cow through a skeletal key point detection algorithm, calculating the angle change between the key joint points in adjacent frames, and obtaining the joint angle change data.

[0030] As an embodiment of the present application, the dynamic gait video is a multi-angle video obtained by setting multiple fixed point high-definition cameras and multiple continuously movable position mobile cameras in the cowshed.

[0031] As an embodiment of the present application, the estrus monitoring and recognition large model includes a posture analysis module, a behavior analysis module, a life state analysis module and an integration module; wherein the posture analysis module receives the dynamic gait feature parameters of the beef cow, analyzes the posture and joint change information of the cow; the behavior analysis module receives the behavior data, analyzes the behavior pattern information of the cow; the life state analysis module receives the life data, analyzes the state of the beef cow according to the beef drinking and water drinking and weight change data; the integration module integrates and analyzes the analysis results of the posture analysis module, the behavior analysis module and the life state analysis module, to avoid the influence of life data on the estrus state of beef.

[0032] As an embodiment of the present application, the pre-processing of the dynamic gait video in S2 further includes an abnormal frame identification and repair step: constructing an abnormal frame detection model through a convolutional neural network (CNN), determining the fuzziness and occlusion rate of the extracted video frames, and marking as an abnormal frame when the fuzziness is > 30% or the occlusion rate is > 40%; using a 3-frame gait feature interpolation algorithm to repair the abnormal frame, generating a supplementary frame conforming to the continuous gait logic, and replacing the original abnormal frame before target tracking.

[0033] Abnormal frame identification (blur + occlusion rate calculation) Blur determination: Calculate image clarity using Laplacian variance method, formula: M x N: Image pixel size (e.g. 1920 x 1080); : Gray value of (x, y) pixel (0-255); : Laplacian operator, used to calculate pixel gray change rate, ; : Blur quantization value (The smaller the value, the more blurred the image) ;Determination threshold: When < 50, blur > 30%, marked as abnormal frame; Occlusion rate determination: Based on YOLOv8 target detection algorithm, calculate the occlusion rate of beef cattle area, formula: : Actual pixel area of beef cattle (through historical data statistics, adult beef cattle in 1080P image ≈20000 pixels); : Detected unoccluded beef cattle pixel area; : Occlusion rate (percentage).

[0034] Determination threshold: When > 40%, marked as abnormal frame; Abnormal frame repair (gait feature interpolation) Use adjacent 3 frames weighted interpolation method to repair abnormal frame, take step length feature as an example, repair formula: : The repaired step length of the kth frame (abnormal frame); , : The measured step length of the k-1, k+1 frames (normal frames); Weighting coefficient "2": Highlight the reference value of the middle frame and reduce the repair error; Application scope: This formula is also applicable to the repair of step frequency, joint angle and other gait features; As an embodiment of the present application, the estrus monitoring recognition large model further comprises an iterative optimization module: the module collects complete data (including gait, behavior, body temperature, and life data) of no less than 50 groups of verified estrus / non-estrus beef cattle per month, and fine-tunes the weight parameters of the posture analysis module and the behavior analysis module through the small batch gradient descent method; at the same time, the model misjudgment cases (such as false estrus misjudgment as estrus) are recorded, the weight distribution rules of the integration module are corrected through attribution analysis, and the interference rate of life data on estrus determination is reduced to below 5%.

[0035] 50 groups of verified data (10 groups of estrus and 40 groups of non-estrus) are collected per month, and the module weight is updated by using the small batch gradient descent method, and the formula is: 、 : updated / previous weight; η: learning rate (0.001, optimized by learning rate decay strategy); L: cross-entropy loss function (measure the deviation between predicted value and true label): N: batch data amount (50 groups); : true label (1=estrus, 0=non-estrus); : model prediction score; Interference rate control The integration module weight is corrected through attribution analysis, and the interference rate calculation formula is: : number of misjudgments caused by life data abnormalities; : total number of misjudgments; Optimization goal: ≤5%.

[0036] The specific workflow is as follows: First, multi-source data collection is carried out, and multiple fixed-point high-definition network cameras are arranged in key areas such as the passageway, resting area and drinking water area of the cowshed, and mobile cameras such as track robots or unmanned aerial vehicles can be equipped to collect dynamic gait videos of beef cattle from multiple angles and continuously. At the same time, each cow is equipped with a smart ear tag with a built-in temperature sensor to detect and wirelessly transmit body temperature data at a frequency of minutes, and then the video is preprocessed and behavior data is extracted. Video frame extraction: the video stream is decomposed into continuous image frames at a certain frame rate; image denoising: image denoising is performed using Gaussian filtering, median filtering and other algorithms to improve image quality; target detection and tracking: using YOLO, DeepSORT and other target detection and tracking algorithms, each beef cattle in the video is identified and assigned a unique ID, and its motion trajectory is tracked, so that the continuous gait image sequence of a single beef cattle is obtained; Abnormal frame identification and repair: the pre-trained convolutional neural network (CNN) model is used to evaluate the quality of each frame, calculate its blur degree (such as image gradient value) and occlusion rate (area ratio of target being occluded); when the blur degree > 30% or the occlusion rate > 40%, the frame is marked as an abnormal frame; the adjacent 3 frames of gait features are interpolated to repair the abnormal frame, generate a supplementary frame that meets the motion logic, replace the original abnormal frame, and ensure the continuity of subsequent analysis; the processed image sequence is associated and labeled with the body temperature data of the same timestamp.

[0037] Through the behavior analysis module, the video is analyzed to extract behavior data such as the number of times of climbing, the number of times of being climbed, the amount of activity, social behavior, etc. Dynamic gait feature parameter extraction: from the high-quality continuous gait image sequence obtained in S2, dynamic gait feature parameters that can quantify the motion state of beef cattle are extracted, mainly including: step length: the distance between two consecutive landing points of the same hoof print; step frequency: the number of steps completed per unit time; step speed: the product of step length and step frequency, reflecting the speed of motion; joint angle change: through the bone key point detection algorithm (such as HRNet), the shoulder, elbow, knee, hip and other key joints of beef cattle are identified, and the angle change sequence of these key joints between adjacent frames is calculated to form the joint angle change data. Estrus cattle often exhibit characteristic postures such as stiff gait and open hind legs, which can be captured through joint angle change; body center of gravity offset: through the position of the key joint, the motion trajectory of the body center of gravity is calculated to analyze its stability. Estrus cattle may exhibit characteristics such as unstable center of gravity and left-right sway due to discomfort; Then the beef cattle life data are recorded through the automatic feeding station, the intelligent weighing bridge, the water quantity sensor and the like equipment, the beef cattle daily record diet data (the forage intake, the foraging time length), the drinking data and the weight change data, form the life data set reflecting its physiological state, further to the estrus monitoring and recognition large model training, construct a multi-module estrus monitoring and recognition large model, the dynamic gait characteristic parameter, the behavior data, the body temperature data and the life data extracted in S2, S3, S4 are as multi-modal input, a large number of marked estrus / non-estrus sample data are used to train the model, the large model is an integrated learning framework, the core module includes: the posture analysis module: receiving the dynamic gait characteristic parameter, focusing on analyzing the microscopic gait anomaly and the joint movement mode of the cattle; the behavior analysis module: receiving the behavior data, focusing on analyzing the macro behavior mode (such as the surge of the climbing behavior); the life state analysis module: receiving the life data, analyzing the diet drinking and the weight change, for judging the cattle health condition, and as an important basis for identifying false estrus (for example, the disease caused appetite reduction and activity anomaly may be misjudged as estrus), and the beef cattle body temperature will rise when estrus, when the beef cattle are sick or due to the large number of beef cattle, the beef cattle feeding and drinking need to be particularly crowded, which also causes the body temperature to rise, and for the less drinking cattle, when the water intake decreases, the body temperature rises, the change of the body temperature will cause the estrus monitoring to be wrong, which will cause the false estrus state, and when the number of beef cattle is large and the feeding is crowded, the climbing action between the beef cattle for feeding but the purpose is to feed and not to estrus, which will also cause the estrus monitoring to be wrong, the present application considers that the life state of the beef cattle has influence on the action behavior and the body temperature, therefore, the life data are added to the monitoring, so that the false estrus judgment is increased on the basis of estrus and non-estrus, the estrus state judgment error is avoided, and the monitoring accuracy is improved. The integration module: receiving the output (such as the characteristic vector or the probability value) of the above three modules, integrating and analyzing through the weighted fusion or the more complex network (such as the full connection layer), finally outputting the comprehensive determination result. The core role of the module is to weigh the contribution of each data source, and avoid the interference of the abnormal situation in the life data on the estrus determination.

[0038] Iterative optimization module: The system has the ability of continuous learning. Every month, the system automatically collects no less than 50 groups of complete data of estrus / non-estrus cattle verified by artificial verification, fine-tunes the model parameters of the posture and behavior analysis module through small batch gradient descent method to adapt to the changes of the cattle herd. At the same time, the system records all misjudgment cases, traces the source of errors through attribution analysis, and accordingly modifies and integrates the weight distribution rules of the module, so as to continuously optimize the model, the goal is to reduce the interference rate of life data and other factors on estrus determination to below 5%; finally, estrus state recognition and output In the actual application stage, the system identifies the identity information of the cattle through intelligent ear tags, and inputs the dynamic gait characteristic parameters, behavior data, body temperature data and life data of the cattle into the trained estrus monitoring and recognition big model in real time or at regular intervals. After the model analyzes and calculates, the recognition result of the current estrus state of the beef cattle is output, the result is divided into three categories: estrus, non-estrus and false estrus. The system can immediately notify the feeders through sound and light alarm, mobile phone APP push, management software interface prompt and other ways.

Claims

1. A method for monitoring the estrus status of beef cattle based on dynamic gait analysis, characterized in that, Includes the following steps: S1: Collect dynamic gait videos of beef cattle and monitor their body temperature in real time to obtain body temperature data; S2: Preprocess the acquired dynamic gait video, including video frame extraction, image denoising, target detection and tracking, to obtain a continuous gait image sequence of a single beef cattle, and label the real-time body temperature data; Dynamic gait videos are analyzed using an analysis module to obtain beef cattle behavior data; S3: Extract dynamic gait feature parameters of beef cattle from the continuous gait image sequence. The dynamic gait feature parameters include stride length, stride frequency, stride speed, joint angle changes, and body center of gravity offset. S4: Obtain living data by recording the diet and weight changes of beef cattle in daily work; S5: Input the extracted dynamic gait feature parameters, behavioral data, body temperature data, and lifestyle data into the estrus monitoring and recognition model for training; S6: Identify cattle identity information through ear tags, input the acquired dynamic gait characteristic parameters, behavioral data, body temperature data and living data into the estrus monitoring and identification model for analysis, and obtain the identification results of the corresponding beef cattle estrus status, including estrus, no estrus and false estrus.

2. The method for monitoring the estrus status of beef cattle based on dynamic gait analysis according to claim 1, characterized in that, The living data mentioned in S4 also includes data on cattle drinking water.

3. The method for monitoring the estrus status of beef cattle based on dynamic gait analysis according to claim 1, characterized in that, The specific process for extracting joint angle changes in S3 is as follows: key joint points of beef cattle are identified through a skeletal key point detection algorithm, the angle changes between key joint points in adjacent frames are calculated, and joint angle change data are obtained.

4. The method for monitoring the estrus status of beef cattle based on dynamic gait analysis according to claim 1, characterized in that, The dynamic gait video is obtained from multiple angles by setting up multiple fixed high-definition cameras and multiple mobile cameras that can be moved continuously in the cattle shed.

5. The method for monitoring the estrus status of beef cattle based on dynamic gait analysis according to claim 1, characterized in that, The estrus monitoring and identification model includes a posture analysis module, a behavior analysis module, a living condition analysis module, and an integration module. The posture analysis module receives dynamic gait characteristic parameters of beef cattle and analyzes the cattle's posture and joint changes. The behavior analysis module receives behavioral data and analyzes the cattle's behavioral patterns. The living condition analysis module receives living data and analyzes the cattle's condition based on their diet, water intake, and weight changes. The integration module combines the analysis results from the posture analysis module, behavior analysis module, and living condition analysis module to avoid the influence of living data on the estrus status of beef cattle.

6. The method for monitoring the estrus status of beef cattle based on dynamic gait analysis according to claim 1, characterized in that, The preprocessing of dynamic gait video in S2 also includes abnormal frame identification and repair steps: an abnormal frame detection model is constructed by a convolutional neural network, and the extracted video frames are judged for blur and occlusion rate. When the blur is greater than 30% or the occlusion rate is greater than 40%, they are marked as abnormal frames. The abnormal frames are repaired by using a gait feature interpolation algorithm of three adjacent frames to generate supplementary frames that conform to the continuous gait logic, and the original abnormal frames are replaced before target tracking is performed.

7. The method for monitoring the estrus status of beef cattle based on dynamic gait analysis according to claim 5, characterized in that, The estrus monitoring and identification model also includes an iterative optimization module: this module collects no less than 50 sets of complete data from verified estrus / non-estrus cattle each month, including dynamic gait characteristic parameters, behavioral data, body temperature data, and living data. It fine-tunes the weight parameters of the posture analysis module and behavior analysis module using the small-batch gradient descent method. At the same time, it records cases of model misjudgment, such as false estrus being misjudged as estrus. It corrects the weight allocation rules of the integration module through attribution analysis, so that the interference rate of living data on estrus determination is reduced to below 5%.

Images