A machine vision-based animal track monitoring method and system

Abstract

The application discloses an animal track monitoring method and system based on machine vision, and relates to the fields of computer vision and image processing.The method comprises the following steps: collecting a first image and a second image through a binocular camera, and pre-processing the first image and the second image; extracting a feature point set from the pre-processed first image and second image, and generating a first image feature descriptor and a second image feature descriptor; obtaining matching points of the first image in the second image based on the first image feature descriptor and the second image feature descriptor; screening corresponding feature points of the first image and the second image based on the matching points of the first image in the second image; and calculating three-dimensional space coordinates of an animal through an adaptive optimization algorithm based on the corresponding feature points of the first image and the second image, and obtaining accurate position information of the animal.Through the three-dimensional coordinate calculation of the adaptive optimization algorithm, the robustness and reliability of the animal track monitoring are improved.

Description

Technical Field

[0001] This invention relates to the fields of computer vision and image processing, and in particular to a method and system for monitoring animal tracks based on machine vision. Background Technology

[0002] In recent years, with the rapid development of machine vision technology, animal tracking monitoring systems based on binocular cameras have gradually become a research hotspot. These systems achieve real-time monitoring and precise tracking of animal behavior by acquiring and processing images captured by binocular cameras. Traditional animal tracking monitoring methods mainly rely on manual observation and recording, which is not only time-consuming and labor-intensive but also easily affected by human factors, resulting in low data accuracy and reliability. With the advancement of computer vision and image processing technologies, machine vision-based animal tracking monitoring systems are gradually demonstrating enormous potential. These systems utilize advanced image processing algorithms to automatically identify and track animal movement trajectories, greatly improving the efficiency and accuracy of monitoring.

[0003] However, existing machine vision-based animal tracking monitoring systems still have some shortcomings. First, existing systems often rely on fixed parameter settings during feature point extraction and matching, lacking adaptability and flexibility. These fixed parameter settings can lead to inaccurate feature point extraction under different environmental and lighting conditions, thus affecting subsequent matching and 3D coordinate calculation. Second, existing systems typically use simple threshold methods in the matching point selection stage, which cannot effectively eliminate false matches, resulting in significant errors in the final 3D coordinate calculation results. These problems limit the application scope and practical effectiveness of existing systems, especially in complex and variable field environments. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a machine vision-based method for monitoring animal tracks to solve the problems of inaccurate feature point extraction and difficulty in eliminating mismatched points.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a machine vision-based method for monitoring animal tracks, comprising: acquiring a first image and a second image using a binocular camera, and preprocessing the first image and the second image; extracting a set of feature points from the preprocessed first image and the second image to generate a first image feature descriptor and a second image feature descriptor; obtaining matching points of the first image in the second image based on the first image feature descriptor and the second image feature descriptor; filtering corresponding feature points of the first image and the second image based on the matching points of the first image in the second image; and calculating the animal's three-dimensional spatial coordinates using an adaptive optimization algorithm based on the corresponding feature points of the first image and the second image to obtain the animal's accurate location information.

[0008] As a preferred embodiment of the machine vision-based animal tracking monitoring method of the present invention, the preprocessing includes size standardization, grayscale conversion, and noise reduction.

[0009] As a preferred embodiment of the machine vision-based animal tracking monitoring method of the present invention, the specific steps for extracting the feature point set from the preprocessed first image and second image are as follows:

[0010] Common datasets containing animal images were collected and labeled. An initial threshold t0 was randomly set, and the FAST algorithm was run based on the initial threshold t0. The threshold t was progressively adjusted, and the detection precision and recall at each threshold t were evaluated. The threshold t with the highest F1 score was selected as the final threshold t. Based on the final threshold t, the thresholds in the FAST algorithm were determined through cross-validation and independent test set validation.

[0011] The FAST algorithm is used to extract feature point sets from the preprocessed first and second images.

[0012] As a preferred embodiment of the machine vision-based animal tracking monitoring method of the present invention, the specific steps for generating a first image feature descriptor and a second image feature descriptor are as follows:

[0013] The feature point set includes multiple feature points, and a neighborhood of a preset size is selected around each feature point.

[0014] The gradient of each neighborhood is obtained using the Sobel algorithm;

[0015] The weighted gradient algorithm is used to calculate the number of first and second image feature descriptors.

[0016] As a preferred embodiment of the machine vision-based animal tracking monitoring method of the present invention, the specific steps for obtaining the matching point of the first image in the second image based on the first image feature descriptor and the second image feature descriptor are as follows:

[0017] The appropriate checks parameter is selected through K-fold cross-validation to balance the accuracy and computational efficiency of the FLANN algorithm;

[0018] The similarity between the first image feature descriptor and the second image feature descriptor is measured using a Euclidean distance metric.

[0019] The FLANN algorithm is used to find matching points between the first image and the second image.

[0020] As a preferred embodiment of the machine vision-based animal tracking monitoring method of the present invention, the specific steps for filtering corresponding feature points in the first image and the second image based on matching points in the first image and the second image are as follows:

[0021] The matching points in the first image and the second image are filtered by comparing the distance ratio between the best match and the second best match of each feature point to select reliable matching points.

[0022] Based on the matching points in the second image from the first image, reliable matching points are selected using a ratio test.

[0023] Based on reliable matching points of feature points in the first and second images, corresponding feature points of the first and second images are obtained.

[0024] As a preferred embodiment of the machine vision-based animal tracking monitoring method of the present invention, the following steps are taken: Based on the corresponding feature points of the first and second images, an adaptive optimization algorithm is used to calculate the animal's three-dimensional spatial coordinates to obtain accurate animal location information.

[0025] The RANSAC algorithm is used to calculate the fundamental matrices of the first and second images based on the corresponding feature points of the first and second images.

[0026] The fundamental matrix is ​​combined with the internal parameters of the binocular camera to convert it into an essential matrix;

[0027] By performing singular value decomposition on the essential matrix, the relative rotation matrix and translation vector between the two views can be recovered.

[0028] Using the recovered rotation and translation information, combined with the coordinates of each matching point in the first and second images, a triangulation algorithm is used to calculate the animal's three-dimensional coordinates.

[0029] The output three-dimensional coordinates of the animal are optimized using an adaptive optimization algorithm to obtain the three-dimensional spatial coordinates of the animal.

[0030] By using the animal's three-dimensional spatial coordinates, we can obtain the animal's accurate location information in three-dimensional space.

[0031] Secondly, the present invention provides an animal tracking monitoring system based on machine vision, comprising: an image acquisition and preprocessing module, a feature point extraction and descriptor generation module, a feature point matching module, a matching point filtering module, and a three-dimensional spatial coordinate calculation module; the image acquisition and preprocessing module is used to acquire a first image and a second image through a binocular camera, and preprocess the first image and the second image; the feature point extraction and descriptor generation module is used to extract a set of feature points from the preprocessed first image and the second image, and generate a first image feature descriptor and a second image feature descriptor; the feature point matching module is used to obtain matching points of the first image in the second image based on the first image feature descriptor and the second image feature descriptor; the matching point filtering module is used to filter corresponding feature points of the first image and the second image based on the matching points of the first image in the second image; and the three-dimensional spatial coordinate calculation module is used to calculate the animal's three-dimensional spatial coordinates based on the corresponding feature points of the first image and the second image using an adaptive optimization algorithm, and obtain accurate animal location information.

[0032] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein the computer program, when executed by the processor, implements any step of the machine vision-based animal tracking monitoring method as described in the first aspect of the present invention.

[0033] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the machine vision-based animal tracking monitoring method as described in the first aspect of the present invention.

[0034] The beneficial effects of this invention are as follows: By using adaptive threshold adjustment for feature point extraction and adaptive optimization algorithm-based three-dimensional coordinate calculation, this invention achieves accurate extraction of feature points and high-precision calculation of animal three-dimensional coordinates under different environmental and lighting conditions, significantly improving the robustness and reliability of animal tracking monitoring, and providing high-quality data support for animal behavior research and ecological monitoring. Attached Figure Description

[0035] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is a flowchart of the machine vision-based animal tracking monitoring method in Example 1.

[0037] Figure 2 This is a system diagram of the machine vision-based animal tracking monitoring system in Example 1. Detailed Implementation

[0038] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0039] Example 1, referring to Figure 1 and Figure 2 This is the first embodiment of the present invention, which provides a machine vision-based animal tracking monitoring method, including the following steps:

[0040] S1. Preprocessing includes size normalization, grayscale conversion, and noise reduction.

[0041] It should be noted that the dimensions of the first and second images are determined, and the images are scaled to the target size using a bilinear interpolation algorithm;

[0042] The standardized color image is converted to grayscale values ​​using a grayscale conversion formula, expressed as follows:

[0043] I g =0.299·r+0.587·g+0.114·b;

[0044] Among them, I g These are grayscale values. r represents the red channel value of a pixel in the color image, g represents the green channel value of a pixel in the color image, and b represents the blue channel value of a pixel in the color image.

[0045] By selecting an appropriate Gaussian kernel size and standard deviation, a Gaussian filtering denoising algorithm is used to perform a convolution operation on the grayscale image to smooth out the noise in the image.

[0046] S2. Extract a set of feature points from the preprocessed first image and second image.

[0047] S2.1. Collect and label common datasets containing animal images. Randomly set an initial threshold t0. Run the FAST algorithm based on the initial threshold t0, progressively adjust the threshold t, and evaluate the detection precision and recall at each threshold t. Select the threshold t with the highest F1 score as the final threshold t. Based on the final threshold t, validate through cross-validation and independent test set to determine the threshold in the FAST algorithm.

[0048] S2.2 Extract feature point sets from the preprocessed first and second images using the FAST algorithm. It should be noted that the expression for extracting the feature point set is:

[0049]

[0050] Where Ge is the set of extracted feature points, (x,y) is the pixel position in the image, and N r (x,y) represents the neighborhood centered at (x,y), and (x',y') represents the neighborhood N. r Let I(x',y') represent the pixel coordinates in the (x,y) table, and let I(x,y) represent the pixel value of the image at (x,y). Let M be a continuous subset on the annulus. This indicates the threshold for brightness difference.

[0051] It should also be noted that the expression is evaluated as follows:

[0052] For each pixel (x, y) in the image, in its neighborhood N r Find a continuous subset M within (x,y) such that the brightness difference |I(x',y')-I(x,y)| between all pixels (x',y') and the center pixel (x,y) in subset M is greater than a threshold. Pixels (x, y) that meet the conditions are selected as feature points and form the feature point set Ge.

[0053] S3. Generate the first image feature descriptor and the second image feature descriptor.

[0054] S3.1 The feature point set includes multiple feature points, and a region of a preset size is selected around the feature points to be set as the neighborhood.

[0055] S3.2 Use the Sobel algorithm to obtain the gradient of each neighborhood.

[0056] Specifically, Sobel convolution kernels are defined in the horizontal and vertical directions, and these two kernels are used to perform convolution operations with the image to calculate the horizontal and vertical gradients of each pixel, ultimately obtaining the gradient of each neighborhood.

[0057] S3.2 Calculate the first and second image feature descriptors using the weighted gradient algorithm. The expression is:

[0058]

[0059] Where H(i) is the first and second image feature descriptors of the i-th feature point, (xl, yl) are the coordinates of the neighboring pixels, and V(p i ) is the i-th feature point p i The surrounding neighborhood, w(x) l ,y l ) represents the weight of pixels within the neighborhood of a feature point. Let α represent the gradient of image I at position (xl, yl), where α is an adjustment factor and C is a constant.

[0060] It should be noted that the weight w(x) of the pixels in the neighborhood of the feature point l ,y l The expression is:

[0061]

[0062] Among them, (x i ,y i ) represents the coordinates of the i-th feature point, σ d is the Gaussian kernel standard deviation.

[0063] S4. Based on the first image feature descriptor and the second image feature descriptor, obtain the matching point of the first image in the second image.

[0064] S4.1. Select appropriate check parameters using K-fold cross-validation to balance the accuracy and computational efficiency of the FLANN algorithm.

[0065] Specifically, a series of candidate check values ​​are defined, and then the dataset is split into multiple subsets for K-fold cross-validation. For each check value, the FLANN algorithm is trained on the training set and its performance is validated on the validation set. The accuracy and computation time are recorded, and finally the check value with the best average performance on all validation sets is selected.

[0066] S4.2 Use Euclidean distance metric to measure the similarity between the first image feature descriptor and the second image feature descriptor.

[0067] Specifically, the Euclidean distance between each pair of feature descriptor vectors is calculated. The smaller the distance value, the more similar the feature descriptors are, thereby determining the matching relationship between feature points.

[0068] S4.3 Find the matching point between the first image and the second image using the FLANN algorithm. The expression is:

[0069]

[0070] Where M(i) is the matching point of the i-th feature point in the first image in the second image, and d(F L,i ,F R,j F is the Euclidean distance between the first image feature descriptor and the second image feature descriptor. L,i F represents the feature descriptor of the i-th feature point in the first image. R,j This represents the feature descriptor of the j-th feature point in the second image, where β is the adjustment coefficient and D... LR The spatial distance is between feature points in the first and second images.

[0071] It should be noted that the Euclidean distance d(F) between the first image feature descriptor and the second image feature descriptor is... L,i ,F R,j The expression is:

[0072]

[0073] Among them, F L,i,k F represents the value of the k-th dimension of the feature descriptor for the i-th feature point in the first image. R,j,k This represents the value of the k-th dimension of the feature descriptor for the i-th feature point in the second image, where n is the number of dimensions of the feature descriptor;

[0074] The spatial distance D between feature points in the first and second images LR The expression is:

[0075]

[0076] Among them, (x L,i ,y L,i (x) represents the coordinates of the i-th feature point in the first image. R,j ,y R,j ) is the coordinate of the j-th feature point in the second image.

[0077] S5. Based on the matching points of the first image in the second image, filter the corresponding feature points of the first image and the second image.

[0078] S5.1. Filter the matching points of the first image in the second image by comparing the distance ratio of the best match and the second best match of each feature point to filter out reliable matching points.

[0079] S5.2. Based on the matching points of the first image in the second image, reliable matching points are selected using a ratio test. The expression is:

[0080]

[0081] Where R(i) is a reliable matching point of feature points in the first image and the second image, d(F L,i ,F R,M(i) ) is the i-th feature point F in the first image. L,i The best matching point F in the second image R,M(i) The distance between them, d(F) L,i ,F R,K(i) ) is the i-th feature point F in the first image. L,i The second best matching point F in the second image R,K(i) The distance between them.

[0082] It should be noted that, This represents the ratio of the best matching distance to the second-best matching distance, and this ratio is used to evaluate the reliability of the match.

[0083] When d(F) L,i ,F R,M(i) ) <d(F L,i ,F R,K(i) This indicates that the best match is reliable;

[0084] When d(F) L,i ,F R,M(i) )≥d(F L,i ,F R,K(i) If the match is not found, then the matching may be unreliable.

[0085] S5.3. Based on reliable matching points of feature points of the first image and the second image, obtain the corresponding feature points of the first image and the second image.

[0086] S6. Based on the corresponding feature points of the first and second images, the animal's three-dimensional spatial coordinates are calculated using an adaptive optimization algorithm to obtain the animal's accurate location information.

[0087] S6.1 Using the RANSAC algorithm, calculate the fundamental matrix of the first and second images based on the corresponding feature points of the first and second images. The expression is:

[0088] Q=({(u L,i ,v L,i ),(u R,i ,v R,i )});

[0089] Where Q is the fundamental matrix of the final output, (u L,i ,v L,i(u) represents the image coordinates of the i-th feature point in the first image. R,i ,v R,i ) represents the image coordinates of the i-th feature point in the second image.

[0090] S6.2 Combine the basic matrix with the internal parameters of the binocular camera to convert it into an essential matrix.

[0091] S6.3 By performing singular value decomposition on the essential matrix, the relative rotation matrix and translation vector between the two views are recovered.

[0092] S6.4. Using the recovered rotation and translation information, combined with the coordinates of each matching point in the first and second images, a triangulation algorithm is used to calculate the animal's three-dimensional coordinates. The expression is:

[0093]

[0094] in, It outputs the animal's three-dimensional coordinates, P L It is the projection matrix of the left camera, P R It is the projection matrix of the right camera.

[0095] It should be noted that the projection matrix P of the left camera L The expression is:

[0096]

[0097] Among them, f x and f y It is the focal length of the left camera (usually in pixels), c x and c y It is the offset of the principal point of the left camera image (usually the coordinates of the image center);

[0098] Projection matrix P of the right camera R The expression is:

[0099]

[0100] Among them, f Rx and f Ry It is the focal length of the right camera, c Rx and c Ry It is the principal point offset of the right camera image, t x t y and t z It is the translation vector of the right camera relative to the left camera.

[0101] S6.5. The output animal 3D coordinates are optimized using an adaptive optimization algorithm to obtain the animal's 3D spatial coordinates, expressed as:

[0102]

[0103] in, This represents the animal's three-dimensional spatial coordinates, where B is the number of corresponding feature points, and w... i For adaptive weights, It is the square of the reprojection error of the left camera. It is the square of the reprojection error of the right camera. It is a three-dimensional world coordinate system.

[0104] It should be noted that the adaptive weight w i The expression is:

[0105]

[0106] Where α is the adjustment coefficient, and A(i) represents the response value of the i-th feature point.

[0107] S6.6 Obtain the animal's accurate location information in three-dimensional space using the animal's three-dimensional spatial coordinates.

[0108] This embodiment also provides an animal tracking monitoring system based on machine vision, including: an image acquisition and preprocessing module, a feature point extraction and descriptor generation module, a feature point matching module, a matching point filtering module, and a three-dimensional spatial coordinate calculation module;

[0109] The image acquisition and preprocessing module is used to acquire a first image and a second image through a binocular camera, and to preprocess the first image and the second image.

[0110] The feature point extraction and descriptor generation module is used to extract a set of feature points from the preprocessed first image and second image, and generate a first image feature descriptor and a second image feature descriptor;

[0111] The feature point matching module is used to obtain matching points of the first image in the second image based on the first image feature descriptor and the second image feature descriptor;

[0112] The matching point filtering module is used to filter corresponding feature points in the first image and the second image based on matching points in the first image.

[0113] The three-dimensional spatial coordinate calculation module is used to calculate the animal's three-dimensional spatial coordinates based on the corresponding feature points of the first and second images, and obtain the animal's accurate location information through an adaptive optimization algorithm.

[0114] This embodiment also provides a computer device applicable to the case of a machine vision-based animal tracking monitoring method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the machine vision-based animal tracking monitoring method proposed in the above embodiment.

[0115] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0116] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the machine vision-based animal tracking monitoring method proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0117] In summary, this invention achieves accurate extraction of feature points and high-precision calculation of animal 3D coordinates under different environmental and lighting conditions by using adaptive threshold adjustment for feature point extraction and adaptive optimization algorithm-based 3D coordinate calculation. This significantly improves the robustness and reliability of animal tracking monitoring and provides high-quality data support for animal behavior research and ecological monitoring.

[0118] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for monitoring animal tracks based on machine vision, characterized in that: include, The first image and the second image are acquired using a binocular camera, and the first image and the second image are preprocessed. Feature point sets are extracted from the preprocessed first and second images to generate a first image feature descriptor and a second image feature descriptor; Based on the first image feature descriptor and the second image feature descriptor, obtain the matching point of the first image in the second image; Based on the matching points in the first image and the second image, filter the corresponding feature points in the first image and the second image; Based on the corresponding feature points of the first and second images, the animal's three-dimensional spatial coordinates are calculated using an adaptive optimization algorithm to obtain the animal's accurate location information; The output animal 3D coordinates are optimized using an adaptive optimization algorithm, resulting in the following expression for the animal's 3D spatial coordinates: ； in, It is the animal's three-dimensional spatial coordinates. For the corresponding number of feature points, For adaptive weights, It is the square of the reprojection error of the left camera. It is the square of the reprojection error of the right camera. It is a three-dimensional world coordinate system; Adaptive weights The expression is: ； in, It is an adjustment factor. It means the first The response value of each feature point.

2. The machine vision-based animal tracking monitoring method as described in claim 1, characterized in that: The preprocessing includes size normalization, grayscale conversion, and noise reduction.

3. The machine vision-based animal tracking monitoring method as described in claim 2, characterized in that: The specific steps for extracting feature point sets from the preprocessed first and second images are as follows: Collect and label common datasets containing animal images, and randomly set initial thresholds. Based on the initial threshold Run the FAST algorithm and gradually adjust the threshold. And evaluate each threshold Based on the detection precision and recall, the threshold with the highest F1 score is selected. As the final threshold Based on the final threshold The threshold in the FAST algorithm was determined through cross-validation and independent test set validation. ; The FAST algorithm is used to extract feature point sets from the preprocessed first and second images.

4. The machine vision-based animal tracking monitoring method as described in claim 3, characterized in that: The specific steps for generating the first and second image feature descriptors are as follows. The feature point set includes multiple feature points, and a neighborhood of a preset size is selected around each feature point. The gradient of each neighborhood is obtained using the Sobel algorithm; The weighted gradient algorithm is used to compute the first and second image feature descriptors.

5. The machine vision-based animal tracking monitoring method as described in claim 4, characterized in that: The specific steps for obtaining the matching point of the first image in the second image based on the first image feature descriptor and the second image feature descriptor are as follows: The appropriate checks parameter is selected through K-fold cross-validation to balance the accuracy and computational efficiency of the FLANN algorithm; The similarity between the first image feature descriptor and the second image feature descriptor is measured using a Euclidean distance metric. The FLANN algorithm is used to find matching points between the first image and the second image.

6. The machine vision-based animal tracking monitoring method as described in claim 5, characterized in that: The specific steps for filtering corresponding feature points in the first and second images based on matching points in the first and second images are as follows. The matching points in the first image and the second image are filtered by comparing the distance ratio between the best match and the second best match of each feature point to select reliable matching points. Based on the matching points in the second image from the first image, reliable matching points are selected using a ratio test. Based on reliable matching points of feature points in the first and second images, corresponding feature points of the first and second images are obtained.

7. The machine vision-based animal tracking monitoring method as described in claim 6, characterized in that: The animal's three-dimensional spatial coordinates are calculated using an adaptive optimization algorithm based on the corresponding feature points of the first and second images to obtain the animal's accurate location information. The specific steps are as follows: The RANSAC algorithm is used to calculate the fundamental matrices of the first and second images based on the corresponding feature points of the first and second images. The fundamental matrix is ​​combined with the internal parameters of the binocular camera to convert it into an essential matrix; By performing singular value decomposition on the essential matrix, the relative rotation matrix and translation vector between the two views can be recovered. Using the recovered rotation and translation information, combined with the coordinates of each matching point in the first and second images, a triangulation algorithm is used to calculate the animal's three-dimensional coordinates. The output three-dimensional coordinates of the animal are optimized using an adaptive optimization algorithm to obtain the three-dimensional spatial coordinates of the animal. By using the animal's three-dimensional spatial coordinates, we can obtain the animal's accurate location information in three-dimensional space.

8. A machine vision-based animal tracking monitoring system, based on the machine vision-based animal tracking monitoring method according to any one of claims 1 to 7, characterized in that: It includes an image acquisition and preprocessing module, a feature point extraction and descriptor generation module, a feature point matching module, a matching point filtering module, and a three-dimensional spatial coordinate calculation module; The image acquisition and preprocessing module is used to acquire a first image and a second image through a binocular camera, and to preprocess the first image and the second image. The feature point extraction and descriptor generation module is used to extract a set of feature points from the preprocessed first image and second image, and generate a first image feature descriptor and a second image feature descriptor; The feature point matching module is used to obtain matching points of the first image in the second image based on the first image feature descriptor and the second image feature descriptor; The matching point filtering module is used to filter corresponding feature points in the first image and the second image based on matching points in the first image. The three-dimensional spatial coordinate calculation module is used to calculate the animal's three-dimensional spatial coordinates based on the corresponding feature points of the first and second images, and obtain the animal's accurate location information through an adaptive optimization algorithm. The output animal 3D coordinates are optimized using an adaptive optimization algorithm, resulting in the following expression for the animal's 3D spatial coordinates: ； in, It is the animal's three-dimensional spatial coordinates. For the corresponding number of feature points, For adaptive weights, It is the square of the reprojection error of the left camera. It is the square of the reprojection error of the right camera. It is a three-dimensional world coordinate system; Adaptive weights The expression is: ； in, It is an adjustment factor. It means the first The response value of each feature point.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the machine vision-based animal tracking monitoring method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the machine vision-based animal tracking monitoring method according to any one of claims 1 to 7.

Images