Cow body measurement method and system based on point cloud segmentation and binocular vision

Abstract

The invention discloses a cattle body measurement method and system based on point cloud segmentation and binocular vision, and realizes non-contact automatic measurement of cattle body size parameters by fusing binocular vision perception and point cloud segmentation technologies. A ZED binocular camera is used for synchronously collecting high-resolution images and depth information, and a PointMama model is combined to automatically position key anatomical feature points of a cattle body, so that stress interference of traditional manual contact measurement on sensitive animals is effectively avoided. Two-dimensional pixel coordinates are mapped to a three-dimensional space through joint calibration of a depth image and internal and external parameter matrixes of a camera, the oblique length main axis direction of a dynamic capture body is analyzed in combination with principal components, the space positioning precision is remarkably improved, and the problem of empirical errors of manual measuring tape winding measurement is solved. On the basis of the real-time reasoning capability of ZEDBox edge computing equipment, lightweight deployment of a point cloud segmentation model is realized, and the high-frequency dynamic monitoring requirement of a farm is met.

Description

Technical Field

[0001] The present invention belongs to the technical field of bovine body measurement, and particularly relates to a bovine body measurement method and system based on point cloud segmentation and binocular vision. Background Art

[0002] The body measurement of water buffalo is an important indicator for evaluating its growth and development, production performance, and health status. Traditionally, these measurements mainly rely on manual use of tools such as tape measures, which is time-consuming and may cause stress reactions to animals. With the development of technology, the application of visual measurement technology in the livestock industry has gradually received attention.

[0003] In the field of beef cattle breeding, the research on intelligent breeding technology has advanced rapidly, including aspects such as individual identification, phenotype acquisition, estrus identification, automated feeding, disease detection, and environmental monitoring. The application of these technologies has improved breeding efficiency and management efficiency, providing support for the intelligent development of the livestock industry.

[0004] In pig breeding, there has been research using deep learning technology to capture pig images through ordinary 2D color cameras and apply convolutional neural networks to predict the weight and body size of pigs. This method does not require constructing feature engineering, can comprehensively extract features, and performs excellently in dealing with noisy data and non-linear problems.

[0005] However, there is relatively little research on the visual measurement of water buffalo body size. Given the research progress in this field for other livestock species, relevant technologies can be borrowed in the future, combined with deep learning and computer vision methods, to develop a non-contact body size measurement system suitable for water buffalo, so as to improve measurement efficiency, reduce stress on animals, and promote the intelligent development of the water buffalo breeding industry. Summary of the Invention

[0006] The purpose of the present invention is to provide a bovine body measurement method and system based on point cloud segmentation and binocular vision to solve the problems raised in the above background art.

[0007] To achieve the above purpose, the present invention adopts the following technical solutions: A bovine body measurement method based on point cloud segmentation and binocular vision includes the following steps: The ZED stereo camera is used to obtain a video stream, and the side images of the cattle body are screened out; the point cloud image segmentation is performed on the side images of the cattle body, and the two-dimensional coordinates of the cattle body measurement points are extracted by using the trained PointMamba model; the two-dimensional coordinates are converted into three-dimensional space coordinates, and the coordinate mapping is performed through the internal and external parameter matrices of the ZED stereo camera; the body diagonal length of the cattle body is calculated according to the three-dimensional space coordinates, and the principal component analysis is used to process the cattle body point cloud data to determine the body diagonal length direction and calculate the three-dimensional Euclidean distance between the two end points; the body height and the withers height of the cattle body are calculated, and by extracting the Z-axis coordinate values of the cattle hoof and the acromion or withers region, the height difference between the two is calculated.

[0008] Preferably, when performing point cloud image segmentation on the side images of the cattle body, the PointMamba model is adopted, and the local and global information is dynamically aggregated through the self-attention mechanism.

[0009] Preferably, the conversion of the two-dimensional coordinates into three-dimensional space coordinates includes: The internal parameter matrix K is represented by the following formula:

[0010] where is the image center coordinate, is the camera focal length.

[0011] Preferably, the calculation of the body diagonal length of the cattle body according to the three-dimensional space coordinates includes: The body diagonal length direction is determined by the eigenvalue decomposition of the covariance matrix Σ. The first principal component v1 corresponds to the largest eigenvalue λ1, and the body diagonal length d is calculated by the following formula: where and are the coordinates of the two end points of the body diagonal length.

[0012] Preferably, the calculation of the body height and the withers height of the cattle body includes: The lowest point of the cattle hoof and the highest point of the acromion The difference in the Z-axis coordinates is the body height, and the formula is: ; where and are the Z-axis coordinate values of the highest point of the acromion and the lowest point of the cattle hoof respectively; The formula for calculating the withers height is ; where is the Z-axis coordinate value of the highest point of the withers.

[0013] Preferably, the calculation of the body height and the withers height of the cattle body further includes: calculating the chest girth and the abdominal girth, specifically including: Using the ellipse fitting method, by extracting the highest point chest_top and the lowest point chest_bottom of the chest or abdominal point cloud, generating mirror coordinate points through symmetric transformation along the Y-axis, and calculating the chest circumference using the following formula: ; Among them, is the major axis radius of the fitted ellipse, is the minor axis radius of the fitted ellipse.

[0014] Preferably, the use of the ZED binocular camera to obtain the video stream includes: Triggering the measurement process through the ear tag reader, used to store the body measurement images and data in the Redis database in real time, and realizing data processing through the ZEDBox edge computing device.

[0015] Preferably, the use of the ZED binocular camera to obtain the video stream further includes: Preloading the point cloud segmentation model to accelerate detection, judging whether the ear tag signal and the cattle body are detected. If detected, perform point cloud segmentation and store the data in the database, and loop until all cattle are detected.

[0016] Preferably, the use of the ZED binocular camera to obtain the video stream further includes: Using the ZEDBox equipped with the NVIDIA Jetson module to process 3D data and image data in real time.

[0017] On the other hand, the present invention proposes a cattle body measurement system based on point cloud segmentation and binocular vision, including: An image acquisition module, used to obtain the video stream using the ZED binocular camera and screen out the side images of the cattle body; A point cloud segmentation module, used to perform point cloud image segmentation on the side images of the cattle body, and extract the two-dimensional coordinates of the cattle body measurement points using the trained PointMamba model; A coordinate conversion module, used to convert the two-dimensional coordinates into three-dimensional space coordinates, and perform coordinate mapping through the internal and external parameter matrices of the ZED binocular camera; A body diagonal length calculation module, used to calculate the body diagonal length of the cattle body according to the three-dimensional space coordinates, process the cattle body point cloud data using principal component analysis, determine the body diagonal length direction and calculate the three-dimensional Euclidean distance between the two end points; A height calculation module, used to calculate the body height and withers height of the cattle body, by extracting the Z-axis coordinate values of the cattle hoof and the shoulder peak or withers area, and calculating the height difference between the two.

[0018] The technical effects and advantages of the present invention: A cattle body measurement method and system based on point cloud segmentation and binocular vision proposed by the present invention has the following advantages compared with the prior art: By integrating binocular vision perception and point cloud segmentation technology, the present invention realizes non-contact automatic measurement of cattle body size parameters. The ZED binocular camera is used to synchronously collect high-resolution images and depth information. Combining with the PointMamba model, the key anatomical feature points of the cattle body are automatically located, effectively avoiding the stress interference of traditional manual contact measurement on sensitive animals. Through the joint calibration of the depth image and the internal and external camera parameter matrices, the two-dimensional pixel coordinates are mapped to the three-dimensional space. Combining with the principal component analysis, the main axis direction of the body slant length is dynamically captured, significantly improving the spatial positioning accuracy. For complex curved surface features such as chest circumference and abdominal circumference, the elliptical fitting algorithm is used to symmetrically match the chest / abdomen point cloud data, and the optimal elliptical circumference is fitted through the mathematical relationship between the major axis and minor axis radii, solving the empirical error problem of manual tape measurement. Based on the real-time inference ability of the ZEDBox edge computing device, the lightweight deployment of the point cloud segmentation model is realized, meeting the high-frequency dynamic monitoring requirements of the farm. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a flowchart of the cattle body measurement method based on point cloud segmentation and binocular vision of the present invention; Figure 2 is a block diagram of the cattle body measurement system based on point cloud segmentation and binocular vision of the present invention; Figure 3 is a schematic diagram of chest circumference fitting of the present invention; Figure 4 is a schematic diagram of the automatic detection principle of cattle body size of the present invention; Figure 5 is a flowchart of the automatic detection of cattle body size of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0021] Embodiment 1 In this embodiment, a cattle body measurement method based on point cloud segmentation and binocular vision is provided, as Figure 1 shown, including the following steps: Use the ZED binocular camera to obtain a video stream and screen out the side images of the cattle body; Perform point cloud image segmentation on the side images of the cattle body, and use the trained PointMamba model to extract the two-dimensional coordinates of the cattle body measurement points; Convert the two-dimensional coordinates into three-dimensional space coordinates and perform coordinate mapping through the internal and external parameter matrices of the ZED binocular camera; Calculate the body diagonal length of the cattle body according to the three-dimensional space coordinates, process the cattle body point cloud data by using principal component analysis, determine the body diagonal length direction and calculate the three-dimensional Euclidean distance between the two end points; Calculate the body height and withers height of the cattle body, and calculate the height difference between the two by extracting the Z-axis coordinate values of the cattle hoof and the acromion or withers area.

[0022] Through the fusion of binocular vision perception and point cloud segmentation technology, the non-contact automatic measurement of cattle body dimension parameters is realized.

[0023] In addition, the ZED binocular camera is used to synchronously collect high-resolution images and depth information, and combined with the PointMamba model to automatically locate the key anatomical feature points of the cattle body, effectively avoiding the stress interference of traditional manual contact measurement on sensitive animals.

[0024] Through the joint calibration of the depth image and the internal and external parameter matrices of the camera, the two-dimensional pixel coordinates are mapped to the three-dimensional space, and combined with principal component analysis to dynamically capture the main axis direction of the body diagonal length, significantly improving the spatial positioning accuracy. For complex curved surface features such as chest circumference and abdominal circumference, the elliptical fitting algorithm is used to symmetrically match the chest / abdomen point cloud data, and the optimal elliptical circumference is fitted through the mathematical relationship between the major axis and minor axis radii, solving the empirical error problem of manual tape measurement. Based on the real-time inference ability of the ZEDBox edge computing device, the lightweight deployment of the point cloud segmentation model is realized, meeting the high-frequency dynamic monitoring requirements of the farm.

[0025] On the other hand, the present invention proposes a cattle body measurement system based on point cloud segmentation and binocular vision, as Figure 2 shown, including: An image acquisition module, configured to use a ZED binocular camera to obtain a video stream and screen out the side images of the cattle body; A point cloud segmentation module, configured to perform point cloud image segmentation on the side image of the cattle body and extract the two-dimensional coordinates of the cattle body measurement points by using the trained PointMamba model; A coordinate conversion module, configured to convert the two-dimensional coordinates into three-dimensional space coordinates and perform coordinate mapping through the internal and external parameter matrices of the ZED binocular camera; A body diagonal length calculation module, configured to calculate the body diagonal length of the cattle body according to the three-dimensional space coordinates, process the cattle body point cloud data by using principal component analysis, determine the body diagonal length direction and calculate the three-dimensional Euclidean distance between the two end points; A height calculation module, configured to calculate the body height and withers height of the cattle body, and calculate the height difference between the two by extracting the Z-axis coordinate values of the cattle hoof and the acromion or withers area.

[0026] In addition, when the above modules are executed, they are also used to implement other steps of the above bovine body measurement method based on point cloud segmentation and binocular vision, such as Figures 3 - 5 as shown below: Dataset collection and processing: When collecting bovine body size data, a ZED binocular camera was used to collect high-resolution videos with depth information, and the images in the video stream were screened. Finally, 400 bovine side image data were selected. This method refers to the Tengchong Betelnut River Buffalo Body Size Measurement Manual, designed the measurement points for Betelnut River buffalo, and performed regional segmentation and annotation on the bovine body images, including key points for buffalo body size measurement and regional segmentation annotation.

[0027] Body size measurement standards: By conducting on-site inspections at the Tengchong Betelnut River Buffalo breeding and production base, communicating and discussing with buffalo breeding experts, understanding the buffalo body size measurement standards, methods, and buffalo living habits, the accuracy of data collection and annotation was ensured. The traditional bovine body size measurement process was visited on-site. By communicating with the measurement staff, the key body size data (unit: centimeter) during the growth process of Betelnut River buffalo were clarified: Withers Height: The vertical distance from the highest point of the withers to the ground.

[0028] Body Length: The distance from the front edge of the scapula to the posterior edge of the ischial tuberosity of the cow.

[0029] Circumference of Chest: The vertical circumference of the body at the posterior edge of the scapula.

[0030] Abdominal Circumference: The vertical circumference at the largest part of the abdomen at the anterior edge of the cross.

[0031] Hip Height: The vertical height from the midpoint of the line connecting the two hip angles of the bovine body to the ground.

[0032] Hardware equipment: This method uses a ZED binocular camera and a ZEDBox for bovine body size detection. The ZED binocular camera is developed by Stereolabs, uses two lenses to capture stereo images and generates a depth map by calculating the disparity, and supports high resolution, high frame rate, and long-distance depth perception. The ZEDBox is an edge computing device designed specifically for running AI and computer vision algorithms of the ZED camera, equipped with an NVIDIA Jetson module, which can process 3D data and image data obtained by the ZED camera in real time. Combining the ZED binocular camera and the ZEDBox enables the real-time operation of the bovine body size detection model on the edge device to achieve real-time detection.

[0033] First, collect bovine body image data, manually annotate the bovine body point cloud segmentation map, and construct a point cloud segmentation training dataset. Use the PointMamba model for point cloud image segmentation to obtain body measurement points, and convert the two-dimensional coordinates of the measurement points into three-dimensional space coordinates. Finally, construct a body measurement algorithm based on point cloud segmentation, combined with an automated measurement strategy, to achieve efficient and accurate measurement of bovine body dimensions on edge devices.

[0034] Measure the body dimensions of water buffalo in three steps. First, use a ZED binocular camera to capture the side image of the water buffalo, and use the trained PointMamba model for point cloud image segmentation. Finally, combine the depth information of the image collected by the camera with the body measurement points, and through a series of measurement methods, calculate the body dimension data of the water buffalo in real time on the ZEDBox edge computing device.

[0035] Bovine body point cloud image segmentation model: PointMamba is a point cloud deep learning model based on the Mamba architecture, designed specifically for 3D point cloud classification, semantic segmentation, and instance segmentation tasks. Compared with traditional point cloud processing methods based on MLP or convolution (such as PointNet++ and KPConv), PointMamba dynamically aggregates local and global information through self-attention mechanism, improves the feature expression ability, and overcomes the limitation of the local receptive field. The model mainly includes three types of modules: PointMambaBlock, TransitionDown, and TransitionUp.

[0036] PointMambaBlock is the core computational unit of the model, mainly used to extract point cloud features. It uses the self-attention mechanism to dynamically aggregate local and global information, improving the feature expression ability.

[0037] The TransitionDown module is mainly used to reduce the number of points, which is equivalent to the pooling operation in a convolutional neural network. It performs feature mapping on the downsampled points through an MLP, increases the feature dimension to compensate for information loss and improve computational efficiency, while retaining the main features.

[0038] The TransitionUp module is responsible for restoring the position information lost during downsampling and is used for feature recovery in point cloud segmentation tasks. It uses nearest neighbor interpolation to propagate the features of the low-resolution point cloud to the high-resolution point cloud, retaining local geometric details to ensure more accurate point cloud segmentation results.

[0039] Three-dimensional coordinate transformation: In this method, after the point cloud image segmentation is completed, it is necessary to use the information in the depth image to determine the exact position of the measurement point in the three-dimensional space. The depth information obtained by the ZED camera is the most accurate when the distance from the buffalo is 2.5m. The ZED binocular camera calculates the depth information of each pixel point through stereo vision. Combining the internal and external parameters of the camera, the pixel coordinates of the measurement point can be converted to the real-world coordinate system. First, the camera needs to be calibrated to obtain the internal and external parameters. The ZED series cameras used in this method have been calibrated, and its internal parameter matrix K is represented by the following formula: , where is the image center coordinate, is the camera focal length.

[0040] For a given pixel coordinate (u, v), the depth value of this pixel is D(u, v). The pixel coordinates can be mapped to the camera coordinate system using the following formula: where is the three-dimensional coordinate of this point in the camera coordinate system.

[0041] To convert the camera coordinate to the real-world coordinate system, the external parameters of the camera need to be considered, that is, the rotation matrix and displacement vector of the camera relative to the world coordinate system. Assuming the rotation matrix of the camera is R and the displacement vector is t, the camera coordinate system can be converted to the world coordinate system using the following formula: .

[0042] Calculation of body diagonal length: In the process of cattle farm breeding management, a tape measure is usually used to measure the body diagonal length of cattle. The tape measure is stretched straight along the side of the cattle from the leading edge of the scapula to the trailing edge of the tuber coxae (ass), and the distance between these two measurement points is recorded. To accurately measure the body diagonal length of cattle, principal component analysis (PCA) is used to process the three-dimensional point cloud data of the cattle body. PCA is a commonly used data dimensionality reduction technique that can find the direction with the largest variance in the data, thereby extracting the main axis of the data, that is, the body diagonal length of the cattle. By performing PCA analysis on the point cloud data of the cattle body, the main axis and corresponding eigenvalues in the three-dimensional space can be obtained. The specific steps are as follows: (1) Preprocessing of point cloud data: Assume that the cattle body point cloud is represented by , where is the three-dimensional coordinate of each point, and N is the total number of points. First, the point cloud data is de-centered to obtain the offset of each point relative to the centroid of the point cloud: , where is the centroid of the point cloud.

[0043] (2) Calculate the covariance matrix: Calculate the covariance matrix Σ of the point cloud: ; The covariance matrix Σ describes the distribution of the point cloud in various directions.

[0044] (3) Eigenvalue decomposition: Perform eigenvalue decomposition on the covariance matrix Σ to obtain the eigenvectors and the corresponding eigenvalues , satisfying: .

[0045] Among them, the eigenvector represents the main axis direction of the data, represents the variance magnitudes in various directions.

[0046] (4) Determine the main axis direction: The main axis direction is the eigenvector corresponding to the largest eigenvalue. The first principal component v1 corresponds to the direction with the largest variance, representing the main axis of the point cloud, i.e., the body diagonal length.

[0047] (5) Calculate the three-dimensional Euclidean distance between two endpoints: After determining the main axis, next calculate the three-dimensional Euclidean distance between the two endpoints along the main axis direction. Assume the two endpoints along the main axis are and , then the three-dimensional Euclidean distance d between them can be calculated by the following formula:

[0048] Among them, and are the coordinates of the two endpoints of the body diagonal length.

[0049] Calculation of body height and cross height: The body height refers to the vertical distance from the cow's hoof (ground) to the withers. To measure the body height of the cow, select the point cloud data of the cow's chest part, and select two key parts, the cow's hoof and the withers, and calculate the height difference between them in the Z-axis direction. Let the lowest point in the cow's hoof area be , and the highest point in the withers area be . Therefore, the body height can be calculated by the following formula: Among them, and are the coordinate values on the Z-axis of the highest point of the withers and the lowest point of the cow's hoof respectively; The calculation method of the cross height is similar to that of the body height, but the selected point cloud area is different. For the calculation of the cross height, select the tail point cloud area. Similarly, obtain the Z-axis coordinate values of the cross area and the cow's hoof area respectively, and the cross height can be calculated by the following formula: Among them, is the coordinate value on the Z-axis of the highest point of the cross.

[0050] Calculation of chest circumference and abdominal circumference: The surveyor usually stands on the side of the cow and wraps the tape measure around the chest from the trailing edge of the cow's shoulder margin (chest_top) to the vertical point at the bottom of the chest (chest_bottom), ensuring that the tape measure is horizontal and close to the skin of the water buffalo, and records the measurement on the tape measure as the chest circumference. The measurement method for the abdominal circumference is similar. Taking the calculation of the chest circumference as an example, this method segments the point cloud data of the cow body, extracts the point cloud of the chest area (chest), and calculates the chest circumference based on this. First, select the highest point chest_top and the lowest point chest_bottom of the chest from the segmented chest (chest) point cloud, then extract multiple coordinate points along the Y-axis direction, and create mirror coordinate points on the other side through symmetric transformation. Then, use the ellipse fitting method to approximate the abdominal circumference curve of the cow body. The calculation of the chest circumference is the same, and the brisket_top and brisket_bottom in the brisket area point cloud can be selected for calculation: , where is the major axis radius of the fitted ellipse, is the minor axis radius of the fitted ellipse.

[0051] Open automated measurement of body dimensions: In the actual measurement scenario of the body dimensions of Binglangjiang water buffalo, in order to achieve open automated measurement of water buffalo body dimension data, a cattle ear tag reader is used to determine when to start the body dimension detection of the cow, associate the body dimension detection data with the cattle, and measure the weight of the water buffalo. Finally, the measurement data is sent to the server and mobile devices.

[0052] The specific process is as follows: Start the detection, pre-load the point cloud segmentation model to speed up the model detection speed, determine whether the cattle ear tag signal is detected and whether the cow body is detected. If the cow body is detected, perform point cloud image segmentation of the cow body, and store the cow body dimension detection image and body dimension data in the Redis database. After the data of one cow is uploaded, determine whether the cattle ear tag signal has changed, that is, whether there is a next cow entering the field. If so, repeat the above cow body dimension detection steps until all water buffalo are detected.

[0053] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, for those skilled in the art, they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A bovine body measurement method based on point cloud segmentation and binocular vision, characterized in that, It includes the following steps: Use a ZED binocular camera to obtain a video stream and filter out the side images of the cattle body; Perform point cloud image segmentation on the side images of the cattle body, and use the trained PointMamba model to extract the two-dimensional coordinates of the cattle body measurement points; Convert the two-dimensional coordinates into three-dimensional space coordinates through coordinate mapping using the internal and external parameter matrices of the ZED binocular camera; Calculate the body diagonal length of the cattle body according to the three-dimensional space coordinates, and use principal component analysis to process the cattle body point cloud data to determine the body diagonal length direction and calculate the three-dimensional Euclidean distance between the two end points; Calculate the body height and withers height of the cattle body by extracting the Z-axis coordinate values of the cattle hoof and the acromion or withers area and calculating the height difference between the two; 2. The bovine body measurement method based on point cloud segmentation and binocular vision according to claim 1, characterized in that, When performing point cloud image segmentation on the side images of the cattle body, use the PointMamba model to dynamically aggregate local and global information through the self-attention mechanism; 3. A bovine body measurement method based on point cloud segmentation and binocular vision according to claim 1, characterized in that Converting the two-dimensional coordinates into three-dimensional space coordinates includes: The internal parameter matrix K is represented by the following formula: ; wherein, is the image center coordinate, is the camera focal length.

4. A bovine body measurement method based on point cloud segmentation and binocular vision according to claim 1, characterized in that, Calculating the body diagonal length of the cattle body according to the three-dimensional space coordinates includes: The body diagonal length direction is determined by the eigenvalue decomposition of the covariance matrix Σ, and the first principal component corresponds to the largest eigenvalue , and the body diagonal length d is calculated by the following formula: Among them, and are the coordinates of the two end points of the body diagonal length.

5. A bovine body measurement method based on point cloud segmentation and binocular vision according to claim 1, characterized in that The calculation of the body height and withers height of the cattle body includes: The lowest point of the cow's hoof and the highest point of the acromion The difference in the Z-axis coordinates is the body height, and the formula is: ; Among them, and are the coordinate values on the Z-axis of the highest point of the acromion and the lowest point of the bovine hoof, respectively; The calculation formula for the height of the cross part is ; where is the coordinate value of the highest point of the cross part on the Z-axis.

6. The bovine body measurement method based on point cloud segmentation and binocular vision according to claim 1, wherein, The calculation of the body height and withers height of the cattle body also includes: calculating the chest girth and abdominal girth, specifically including: Using the ellipse fitting method, by extracting the highest point chest_top and the lowest point chest_bottom of the chest or abdominal point cloud, generating mirror coordinate points through symmetric transformation along the Y-axis, and calculating the chest circumference using the following formula: ; Among them, is the major axis radius of the fitted ellipse, is the minor axis radius of the fitted ellipse.

7. A bovine body measurement method based on point cloud segmentation and binocular vision according to claim 1, characterized in that The use of the ZED binocular camera to obtain a video stream includes: Trigger the measurement process through a cattle ear tag reader, used to store the body measurement images and data in the Redis database in real time, and implement data processing through the ZEDBox edge computing device; 8. A bovine body measurement method based on point cloud segmentation and binocular vision according to claim 1, characterized in that, The use of the ZED binocular camera to obtain a video stream also includes: Pre-load the point cloud segmentation model to accelerate detection, determine whether a cattle ear tag signal and a cattle body are detected, if detected, perform point cloud segmentation, and store the data in the database, and loop until all cattle are detected; 9. A bovine body measurement method based on point cloud segmentation and binocular vision according to claim 1, characterized in that The use of the ZED binocular camera to obtain a video stream also includes: Use the ZEDBox to carry the NVIDIA Jetson module to process 3D data and image data in real time; 10. A bovine body measurement system based on point cloud segmentation and binocular vision for implementing the method according to any one of claims 1-9, characterized in that, It includes: An image acquisition module, used to use a ZED binocular camera to obtain a video stream and filter out the side images of the cattle body; A point cloud segmentation module, used to perform point cloud image segmentation on the side images of the cattle body, and use the trained PointMamba model to extract the two-dimensional coordinates of the cattle body measurement points; A coordinate conversion module, used to convert the two-dimensional coordinates into three-dimensional space coordinates through coordinate mapping using the internal and external parameter matrices of the ZED binocular camera; A body diagonal length calculation module, used to calculate the body diagonal length of the cattle body according to the three-dimensional space coordinates, and use principal component analysis to process the cattle body point cloud data to determine the body diagonal length direction and calculate the three-dimensional Euclidean distance between the two end points; A height calculation module, used to calculate the body height and withers height of the cattle body by extracting the Z-axis coordinate values of the cattle hoof and the acromion or withers area and calculating the height difference between the two;

Images