Mutton sheep body size measuring method and system based on machine vision

Abstract

The invention relates to the technical field of mutton sheep body size measurement, and discloses a mutton sheep body size measurement method and system based on machine vision, and the method comprises the steps: obtaining an original side image of a mutton sheep, processing the original side image to obtain a standard mutton sheep image, calculating a contour coefficient, and carrying out the clustering to obtain a clustering center; selecting skin pixels according to the clustering center, and performing expansion operation on the skin pixels to obtain a skin mask; contour extraction is carried out on the epidermis mask, and the outer contour mass center and the outer contour main axis direction of the epidermis area are identified; dynamically adjusting the interception window by taking the mass center of the outer contour as the center and combining the main axis direction of the outer contour, so that the interception window is aligned with the body trend of the mutton sheep, and intercepting to obtain a target side profile view of the mutton sheep; a Hough straight line detection algorithm is adopted to detect and obtain multiple initial straight lines, pruning is carried out to obtain endpoint coordinates of the spine characteristic curve, and mutton sheep body size parameters are calculated and obtained according to the endpoint coordinates. According to the invention, the accuracy and efficiency of mutton sheep body size measurement can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of body size measurement of meat sheep, and particularly to a method and system for measuring the body size of meat sheep based on machine vision. Background Art

[0002] The body size parameters of meat sheep mainly include body length, body height, body width and chest circumference. The body size information can not only intuitively reflect the growth status of meat sheep, but also be an important parameter for estimating indicators such as the lean meat rate and backfat thickness of the carcass of meat sheep. The measurement of the body size of meat sheep is the basis for meat sheep breeding and farming. However, at present, manual measurement is still mostly used in production.

[0003] In an existing technology, by using the method of manual calibration, traditional measuring tools such as rulers and calipers are used for measurement, and the body shape characteristics of meat sheep are analyzed. By measuring and analyzing the images of different parts of the sheep body, corresponding data such as body length and body height are obtained. This method requires the operator to have certain experience and skills. At the same time, due to the frequent movement of meat sheep during the measurement process, the measurement is often affected by factors such as the movement of the sheep body and environmental interference, resulting in inaccurate data and deviation of the measurement results. Therefore, the method of manual calibration has a large error and low accuracy.

[0004] The existing judgment methods for the growth of meat sheep mainly rely on traditional methods based on manual measurement. They not only have a large labor intensity, are time-consuming and laborious in the process, but also are affected by subjectivity, and the accuracy of the detection results is not high. Summary of the Invention

[0005] The present invention provides a method and system for measuring the body size of meat sheep based on machine vision to improve the accuracy and efficiency of meat sheep body size measurement.

[0006] In a first aspect, to solve the above technical problems, the present invention provides a method for measuring the body size of meat sheep based on machine vision, including:

[0007] Obtaining an original side image of a meat sheep;

[0008] Performing standardization processing on the original side image according to a preset aspect ratio to obtain a standard meat sheep image;

[0009] Calculating a contour coefficient according to the standard meat sheep image, determining the number of clustering centers according to the contour coefficient, and performing clustering according to the standard meat sheep image and the number of clustering centers to obtain clustering centers;

[0010] Selecting an epidermal pixel range according to the clustering centers to obtain epidermal pixels, and performing dilation operation on the epidermal pixels to obtain an epidermal mask;

[0011] Extract the contour of the epidermal mask, and use the contour detection algorithm to identify the centroid of the outer contour and the direction of the major axis of the outer contour of the epidermal region; within the preset length-width ratio range of the side profile of the mutton sheep, with the centroid of the outer contour as the center, dynamically adjust the intercept window in combination with the direction of the major axis of the outer contour to align the intercept window with the body orientation of the mutton sheep, and intercept the target side profile view of the mutton sheep;

[0012] Detect the target side profile view at preset angle intervals using the Hough line detection algorithm to obtain multiple initial lines, and perform pruning operations on the multiple initial lines, and then fit the multiple segmented lines to obtain the spine feature curve;

[0013] Determine the endpoint coordinates of the spine feature curve, and calculate the body size parameters of the mutton sheep based on the endpoint coordinates.

[0014] In an alternative embodiment, obtaining the original side image of the mutton sheep includes:

[0015] Use the acquisition device to obtain the side image of the mutton sheep, and send the image data to the data center of the system server;

[0016] According to the side image of the mutton sheep, determine whether the image resolution is greater than the preset resolution. If not, discard the image; if so, use the bicubic interpolation method to adjust the image resolution to the preset resolution to obtain the initial side image of the mutton sheep;

[0017] According to the initial side image of the mutton sheep, determine whether the image file size is greater than the preset size threshold. If so, perform image compression; if not, directly obtain the original side image of the mutton sheep.

[0018] In an alternative embodiment, calculate the silhouette coefficient according to the standard mutton sheep image, determine the number of cluster centers according to the silhouette coefficient, and perform clustering according to the standard mutton sheep image and the number of cluster centers to obtain the cluster centers, including:

[0019] According to the standard mutton sheep image, calculate the silhouette coefficient through the following formula:

[0020]

[0021] Among them, a(i) represents the average distance from sample i to other samples in the same cluster, b(i) represents the minimum distance from sample i to other samples in the same cluster, and s(i) is the silhouette coefficient;

[0022] Traverse different K values, calculate the average value of the silhouette coefficients of all samples, and select the K value that makes the average value of the silhouette coefficients the largest as the number of cluster centers;

[0023] Cluster the standard mutton sheep images according to the number of cluster centers to obtain cluster centers.

[0024] In an alternative embodiment, according to the cluster centers, select the epidermal pixel range to obtain epidermal pixels, and perform a dilation operation on the epidermal pixels to obtain an epidermal mask, including:

[0025] According to the cluster centers, calculate the pixel point density with each cluster center as the core.

[0026] According to the pixel point density, select the epidermal pixel range to obtain epidermal pixels.

[0027] Perform a dilation operation on the epidermal pixels according to the cluster centers to obtain an epidermal mask.

[0028] In an alternative embodiment, the performing a dilation operation on the epidermal pixels according to the cluster centers to obtain an epidermal mask includes:

[0029] Divide the preset small regions according to the epidermal pixels and calculate the regional entropy value.

[0030] According to the regional entropy value and the small regions, determine whether each small region is higher than a preset entropy threshold value. If so, expand the skin color range of the region; if not, do nothing to obtain optimized epidermal pixels.

[0031] According to the optimized epidermal pixels, apply a morphological dilation operation to connect the scattered epidermal pixel points into a complete region to obtain an initial epidermal mask.

[0032] Perform preprocessing on the initial epidermal mask to obtain an epidermal mask.

[0033] In an alternative embodiment, the extracting the contour of the epidermal mask and using a contour detection algorithm to identify the centroid of the outer contour and the direction of the major axis of the outer contour of the epidermal region includes:

[0034] Convert the epidermal mask into a grayscale image and perform binarization processing to obtain a binary image.

[0035] Use a contour detection algorithm to scan the binary image to find the closed boundary of the epidermal region and form a polygonal contour.

[0036] Calculate the geometric center of the polygonal contour through image moments to obtain the centroid of the outer contour.

[0037] Use a minimum area rectangle to fit the polygonal contour, and take the long side direction of the minimum area rectangle as the direction of the major axis of the outer contour.

[0038] In an alternative embodiment, detecting multiple initial lines from the target side profile view at preset angle intervals using the Hough line detection algorithm, and pruning the multiple initial lines to obtain a spine feature curve, includes:

[0039] Scanning the target side profile view at preset angle intervals, and extracting detection points in the image;

[0040] Mapping the detection points to the parameter space, and voting for each possible line parameter through an accumulator;

[0041] Finding the maximum value of the voting results in the accumulator, and connecting the detection points corresponding to the maximum value as the detected initial line;

[0042] Eliminating the lines in the initial lines with lengths less than a preset length threshold, and merging the lines in the initial lines with slope similarities greater than a preset similarity threshold to obtain the pruned lines;

[0043] Selecting multiple lines consistent with the spine trend from the pruned lines, and fitting to obtain a spine feature curve.

[0044] In an alternative embodiment, calculating the body size parameters of the meat sheep based on the endpoint coordinates, includes:

[0045] The relationship between the body size parameters of the meat sheep and the endpoint coordinates is expressed as:

[0046]

[0047] In the formula, S represents the body size parameters of the meat sheep, i.e., the estimated body length of the meat sheep; k represents the proportionality coefficient, m represents the number of spine feature curves, L i represents the length of the i-th line segment, α i represents the angle between the i-th line segment and the horizontal direction; (x1, y1),..., (x i , y i ), (x i+1 , y i+1 ) are the endpoint coordinates of the spine feature curve in sequence.

[0048] In a second aspect, the present invention provides a meat sheep body size measurement system based on machine vision, including:

[0049] An image acquisition module, configured to acquire an original side image of the meat sheep;

[0050] An image normalization module, configured to normalize the original side image according to a preset aspect ratio to obtain a standard meat sheep image;

[0051] A clustering module, configured to calculate the silhouette coefficient based on the standard mutton sheep images, determine the number of cluster centers according to the silhouette coefficient, and perform clustering based on the standard mutton sheep images and the number of cluster centers to obtain cluster centers;

[0052] A mask acquisition module, configured to select the epidermal pixel range according to the cluster centers to obtain epidermal pixels, and perform dilation operation on the epidermal pixels to obtain an epidermal mask;

[0053] An image cropping module, configured to extract the contour of the epidermal mask, and use a contour detection algorithm to identify the centroid of the outer contour and the direction of the major axis of the outer contour of the epidermal region; within a preset length-width ratio range of the side profile of the mutton sheep, with the centroid of the outer contour as the center, dynamically adjust the cropping window in combination with the direction of the major axis of the outer contour to align the cropping window with the body orientation of the mutton sheep, and crop to obtain the target side profile view of the mutton sheep;

[0054] A straight line detection module, configured to detect a plurality of initial straight lines from the target side profile view at preset angle intervals by using a Hough straight line detection algorithm, and perform pruning operation on the plurality of initial straight lines to obtain a spine feature curve;

[0055] A body measurement calculation module, configured to determine the endpoint coordinates of the spine feature curve, and calculate the body measurement parameters of the mutton sheep according to the endpoint coordinates.

[0056] In a third aspect, the present invention further provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the method for measuring the body size of mutton sheep based on machine vision described in any one of the above is implemented.

[0057] In a fourth aspect, the present invention further provides a computer-readable storage medium, which includes a stored computer program. When the computer program runs, it controls the device where the computer-readable storage medium is located to execute the method for measuring the body size of mutton sheep based on machine vision described in any one of the above.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] The present invention relates to the technical field of body measurement of meat sheep, and discloses a method and system for measuring the body size of meat sheep based on machine vision. The method includes obtaining an original side image of a meat sheep; performing standardization processing on the original side image according to a preset length-width ratio to obtain a standard meat sheep image; calculating a contour coefficient according to the standard meat sheep image, determining the number of cluster centers according to the contour coefficient, and performing clustering according to the standard meat sheep image and the number of cluster centers to obtain cluster centers; selecting an epidermal pixel range according to the cluster centers to obtain epidermal pixels, and performing a dilation operation on the epidermal pixels to obtain an epidermal mask; extracting the contour of the epidermal mask, and using a contour detection algorithm to identify the centroid of the outer contour and the direction of the outer contour main axis of the epidermal region; within a preset length-width ratio range of the side profile of the meat sheep, with the centroid of the outer contour as the center, dynamically adjust the intercept window in combination with the direction of the outer contour main axis to align the intercept window with the body orientation of the meat sheep, and intercept to obtain a target side profile view of the meat sheep; at preset angle intervals, use the Hough line detection algorithm to detect the target side profile view to obtain a plurality of initial lines, and perform pruning operations on the plurality of initial lines to obtain a spine feature curve; determine the endpoint coordinates of the spine feature curve, and calculate the body size parameters of the meat sheep according to the endpoint coordinates.

[0060] Through a series of image processing technologies, the present invention provides an efficient and accurate automated measurement solution. First, an original side image of a meat sheep is obtained and standardized to obtain a standard meat sheep image, avoiding errors caused by differences in shooting angles and image quality. Then, an image clustering algorithm is used to extract the epidermal region of the meat sheep, and a dilation operation is performed on the epidermal pixels to further optimize the image quality and ensure the integrity of the shape features. Next, the standard meat sheep image is intercepted through the epidermal mask to obtain a side profile view of the target meat sheep, providing key data for subsequent body size measurement. On this basis, a line detection algorithm is applied to identify the body shape feature lines of the meat sheep, and an effective line integration is performed to obtain a spine feature curve, and then the body length parameter of the meat sheep is accurately calculated in combination with the endpoint coordinates of the spine feature curve. The present invention avoids the errors of traditional manual measurement or relying on simple algorithms, improves the accuracy of meat sheep body size measurement, and at the same time can improve the measurement efficiency compared with manual measurement. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 is a schematic flowchart of a method for measuring the body size of meat sheep based on machine vision provided in the first embodiment of the present invention;

[0062] Figure 2 is a schematic structural diagram of a system for measuring the body size of meat sheep based on machine vision provided in the second embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0063] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to 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. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0064] Referring to Figure 1 , the first embodiment of the present invention provides a method for measuring the body size of meat sheep based on machine vision, including the following steps:

[0065] S11, obtaining the original side image of the meat sheep;

[0066] S12, performing standardization processing on the original side image according to a preset length-width ratio to obtain a standard meat sheep image;

[0067] S13, calculating the silhouette coefficient according to the standard meat sheep image, determining the number of cluster centers according to the silhouette coefficient, and performing clustering according to the standard meat sheep image and the number of cluster centers to obtain the cluster centers;

[0068] S14, selecting the epidermal pixel range according to the cluster centers to obtain epidermal pixels, and performing dilation operation on the epidermal pixels to obtain an epidermal mask;

[0069] S15, performing contour extraction on the epidermal mask, and using a contour detection algorithm to identify the centroid of the outer contour and the direction of the main axis of the outer contour of the epidermal region; within the preset length-width ratio range of the side profile of the meat sheep, with the centroid of the outer contour as the center, dynamically adjusting the intercept window in combination with the direction of the main axis of the outer contour to align the intercept window with the body direction of the meat sheep, and intercepting to obtain the target side profile view of the meat sheep;

[0070] S16, detecting the target side profile view at preset angle intervals by using the Hough line detection algorithm to obtain a plurality of initial lines, and performing pruning operations on the plurality of initial lines, and then fitting a plurality of segmented lines to obtain a spine feature curve;

[0071] S17, determining the endpoint coordinates of the spine feature curve, and calculating the body size parameters of the meat sheep according to the endpoint coordinates.

[0072] Through a series of image processing techniques, the present invention provides an efficient and accurate automated measurement solution. First, the original side image of the mutton sheep is obtained and standardized to obtain a standard mutton sheep image, avoiding errors caused by differences in shooting angles and image quality. Then, the epidermis region of the mutton sheep is extracted using an image clustering algorithm, and the epidermis pixels are dilated to further optimize the image quality and ensure the integrity of the shape features. Next, the standard mutton sheep image is intercepted through an epidermis mask to obtain a side sectional view of the target mutton sheep, providing key data for subsequent body size measurement. On this basis, a straight line detection algorithm is applied to identify the body shape feature lines of the mutton sheep, and the spine feature curve is obtained through effective straight line integration. Furthermore, the body length parameter of the mutton sheep is accurately calculated by combining the endpoint coordinates of the spine feature curve. The present invention avoids the errors of traditional manual measurement or relying on simple algorithms, improves the accuracy of mutton sheep body size measurement, and at the same time can improve the measurement efficiency compared with manual measurement.

[0073] In step S11, it is necessary to obtain the original side image of the mutton sheep.

[0074] In one implementation, obtaining the original side image of the mutton sheep includes:

[0075] Using a collection device to obtain the side image of the mutton sheep and sending the image data to the data center of the system server;

[0076] According to the side image of the mutton sheep, determine whether the image resolution is greater than a preset resolution. If not, discard the image; if so, use bicubic interpolation to adjust the image resolution to the preset resolution to obtain the initial side image of the mutton sheep;

[0077] According to the initial side image of the mutton sheep, determine whether the image file size is greater than a preset size threshold. If so, perform image compression; if not, directly obtain the original side image of the mutton sheep.

[0078] It should be noted that the preset image resolution in the present invention is 1024*1024. The basis for this resolution standard selection is to ensure that the image is clear enough to accurately reflect various body characteristics of the mutton sheep, such as body shape, epidermis condition, muscle distribution, etc. This resolution can maintain image details while not causing the image file to be too large due to excessive resolution, affecting storage and transmission efficiency. In practical applications, a resolution of 1024×1024 is sufficient to support image processing, feature extraction, etc. tasks in subsequent analysis and will not waste too much computing resources. Among them, the bicubic interpolation method is used to adjust the image resolution to the preset resolution, and the image quality is evaluated through the structural similarity index (SSIM). If SSIM≥0.95, it is retained; otherwise, the image is re-collected.

[0079] In step S12, the original side image is normalized according to a preset aspect ratio of length to width to obtain a standard mutton sheep image.

[0080] In one implementation, normalizing the original side image according to a preset aspect ratio of length to width to obtain a standard mutton sheep image includes:

[0081] Performing channel separation on the original side image of the mutton sheep to obtain three separated color channels;

[0082] According to the three separated color channels, performing normalization processing on each color channel to obtain the normalized color channel values;

[0083] Merging the normalized color channel values, and cropping the merged image according to a preset aspect ratio of length to width to obtain a standard mutton sheep image.

[0084] It should be noted that the original side image of the mutton sheep is an RGB color image. Through the channel separation operation, the image can be decomposed into three independent grayscale images: the red channel (R), the green channel (G), and the blue channel (B). The role of channel separation is to process the pixel values of different channels respectively, so as to improve the accuracy of normalization.

[0085] It should be noted that the normalization processing performed in step S12 is to ensure that the input image can meet the requirements of subsequent analysis and processing, and improve the accuracy and consistency of processing. Specifically, the goal of the normalization processing is to convert the original side image of the mutton sheep into a unified standard image format, so that factors such as the aspect ratio, color information, contrast, and brightness of the image can be analyzed on the same scale. Among them, the method of adaptive aspect ratio adjustment and background filling is adopted to maintain the original aspect ratio of the mutton sheep, and the image is adjusted to the standard size by adding background filling (such as black / white areas) to avoid stretching or compression. When normalizing the initial side image of the mutton sheep, first calculate its original aspect ratio. If it is inconsistent with the aspect ratio of the preset standard size (such as 1024×1024), a background filling area (such as black pixels) is added to the edge of the image to expand the image to the standard size while retaining the actual ratio of the mutton sheep.

[0086] In step S13, calculate the silhouette coefficient according to the standard mutton sheep image, determine the number of cluster centers according to the silhouette coefficient, and perform clustering according to the standard mutton sheep image and the number of cluster centers to obtain the cluster centers.

[0087] In one implementation, calculating the silhouette coefficient according to the standard mutton sheep image, determining the number of cluster centers according to the silhouette coefficient, and performing clustering according to the standard mutton sheep image and the number of cluster centers to obtain the cluster centers includes:

[0088] According to the standard mutton sheep images, calculate the silhouette coefficient through the following formula:

[0089]

[0090] where a(i) represents the average distance from sample i to other samples in the same cluster, b(i) represents the minimum distance from sample i to other samples in the same cluster, and s(i) is the silhouette coefficient;

[0091] Traverse different values of K, calculate the average value of the silhouette coefficients of all samples, and select the value of K that makes the average value of the silhouette coefficients the largest as the number of clustering centers. Herein, the present invention traverses the values of K from 2 to 5;

[0092] According to the number of clustering centers, perform clustering on the standard mutton sheep images to obtain clustering centers.

[0093] It should be noted that the silhouette coefficient is an important indicator for measuring the adaptability of samples in the clustering results. A silhouette coefficient greater than 0 indicates that the sample is closer to its belonging clustering center, and the clustering effect is better; a silhouette coefficient close to 0 indicates that the sample is at the boundary between two clusters, and the clustering effect is average; a silhouette coefficient less than zero indicates that the sample is closer to other clustering centers and may be misallocated. By calculating the average value of the silhouette coefficients of all samples, the overall clustering effect can be evaluated, and the optimal number of clustering centers can be selected.

[0094] It is worth noting that the clustering centers represent the aggregation points of color and spatial features in the standard mutton sheep images. These center points are important bases for subsequent epidermal pixel recognition, and the color and position features of each clustering center can be used to define the range of epidermal pixels.

[0095] In step S14, according to the clustering centers, select the epidermal pixel range to obtain epidermal pixels, and perform a dilation operation on the epidermal pixels to obtain an epidermal mask.

[0096] In one implementation, according to the clustering centers, select the epidermal pixel range to obtain epidermal pixels, and perform a dilation operation on the epidermal pixels to obtain an epidermal mask, including:

[0097] S141. According to the clustering centers, take each clustering center as the core and calculate the pixel point density;

[0098] S142. According to the pixel point density, select the epidermal pixel range to obtain epidermal pixels;

[0099] S143. Perform a dilation operation on the epidermal pixels according to the clustering centers to obtain an epidermal mask.

[0100] In step S141, calculate the pixel point density according to the following formula:

[0101]

[0102] Among them, s i (x, y) represents the distance from the current pixel to the clustering center i, and ρ r,j represents the j-th color channel within the radius r, σ is a preset parameter used to control the influence weight of the distance, and p(x, y) represents the pixel point density.

[0103] It should be noted that the size of the radius determines the range used for calculating the density. A smaller radius can accurately capture local density changes but ignores large-scale features, while a larger radius is suitable for handling scenarios with a relatively uniform density distribution. For example, the radius adopted in the present invention is 5.

[0104] In step S142, the selection of epidermal pixels is based on the judgment of the density threshold. By setting the density threshold, the pixel points that satisfy the density greater than the threshold are selected as epidermal pixels. In the present invention, the density threshold is taken as 0.6, which can be further adjusted according to the data distribution.

[0105] It should be noted that the range of epidermal pixels is not only related to the density but also needs to be screened in combination with the color characteristics of the clustering result. By limiting the color range of epidermal pixels in the RGB or HSV color space. For example, in the present invention, the brightness value range of the skin color is [0.3, 0.7].

[0106] In step S143, performing a dilation operation according to the epidermal pixels to obtain an epidermal mask includes:

[0107] Dividing a preset small area according to the epidermal pixels and calculating the regional entropy value;

[0108] Judging whether each small area is higher than a preset entropy value threshold according to the regional entropy value and the small area. If so, the skin color range of this area is expanded. If not, no operation is performed to obtain optimized epidermal pixels;

[0109] Applying a morphological dilation operation according to the optimized epidermal pixels to connect the scattered epidermal pixel points into a complete area to obtain an initial epidermal mask;

[0110] In one implementation, preprocessing the initial epidermal mask to obtain an epidermal mask.

[0111] The preprocessing of the initial epidermal mask to obtain an epidermal mask includes:

[0112] Removing the patches caused by uneven brightness in the initial epidermal mask to obtain a uniform epidermal mask;

[0113] Denoise the uniform skin mask using median filtering to obtain the denoised skin mask;

[0114] Normalize the denoised skin mask to obtain the skin mask.

[0115] It should be noted that in the skin mask, due to light changes, image noise, or environmental interference, patches with uneven brightness may appear. By applying region connection analysis or threshold processing, these patches can be removed to retain a coherent and uniform skin area. This process helps to remove interference caused by environmental factors, thereby improving the quality of the skin mask.

[0116] It should be noted that median filtering is a classic image denoising method that can effectively remove salt-and-pepper noise or small-scale noise points. It replaces the current pixel value by calculating the median within the neighborhood of each pixel, thereby retaining important structural features while smoothing the noise. This is very important for maintaining the smoothness and coherence of the skin mask.

[0117] It is worth noting that the preprocessed skin mask is more stable, can accurately represent the skin area of the mutton sheep, reduce errors and improve the accuracy of subsequent analysis. Especially in the process of contour extraction, the skin mask after these preprocessing steps can effectively improve the efficiency and effect of image processing.

[0118] It should be noted that the regional entropy value is an index used to measure the complexity of pixel distribution in an image, which can reflect the change of pixel values in a small area. In the processing of skin pixels, the high or low of the regional entropy value can distinguish the parts with complex textures in the image (such as background, hair) from the parts with relatively uniform textures (such as skin). By setting the entropy threshold, areas that are more in line with the skin characteristics can be effectively screened out, and noise points or background areas misjudged as skin can be excluded. The setting of the entropy threshold needs to be adjusted according to the characteristics of the actual image (such as resolution, color distribution, etc.). A lower threshold results in background noise being misidentified as skin, while a higher threshold misses some real skin areas. Exemplarily, the entropy threshold adopted in the present invention is 0.8.

[0119] It should be noted that when calculating the entropy value of skin pixels, the image is divided into multiple small regions. The size of these small regions directly affects the processing accuracy and efficiency. If the region is too large, skin and non-skin pixels will be mixed, reducing the recognition accuracy. If the region is too small, more computational overhead will be introduced, affecting the overall efficiency. Exemplarily, the region adopted in the present invention is a 3*3 small region.

[0120] It should be noted that the dilation operation is a basic morphological processing method, and its main function is to expand the neighborhood range of pixel points, thereby connecting scattered pixel points and filling holes in the epidermal area. In this step, the dilation operation can connect the optimized epidermal pixels into a continuous area to form a complete epidermal mask. This step can eliminate the problem of discontinuous epidermal areas caused by noise or other interferences.

[0121] In step S15, contour extraction is performed on the epidermal mask, and the centroid of the outer contour and the direction of the major axis of the outer contour of the epidermal area are identified using a contour detection algorithm; within the preset length-width ratio range of the side profile of the meat sheep, with the centroid of the outer contour as the center and in combination with the direction of the major axis of the outer contour, the cropping window is dynamically adjusted to align the cropping window with the body orientation of the meat sheep, and a target side profile view of the meat sheep is obtained by cropping.

[0122] Among them, the performing contour extraction on the epidermal mask and identifying the centroid of the outer contour and the direction of the major axis of the outer contour of the epidermal area using a contour detection algorithm includes:

[0123] Convert the epidermal mask into a grayscale image and perform binarization processing to obtain a binary image;

[0124] Use a contour detection algorithm to scan the binary image to find the closed boundary of the epidermal area and form a polygon contour;

[0125] Calculate the geometric center of the polygon contour through image moments to obtain the centroid of the outer contour;

[0126] Use the minimum area rectangle to fit the polygon contour, and take the long side direction of the minimum area rectangle as the direction of the major axis of the outer contour.

[0127] Specifically, convert the epidermal mask into a grayscale image and perform binarization processing. For example, use the threshold segmentation method to convert the image into a binary image. Then, the cv2.findContours function in OpenCV can be used to extract the contour, find the closed boundary of the epidermal area, and form a polygon contour. The image moments (Moments) are used to calculate the centroid of the extracted contour. Image moments (Moments) is a statistical method based on pixel distribution and is used to describe the geometric characteristics of the image. The abscissa Cx and ordinate Cy of the centroid are calculated by the following formulas respectively:

[0128]

[0129] where M 00 is the total area of the contour, M 10 and M 01 are the components of the image moments.

[0130] Then, the minimum area rectangle is used to fit the contour to obtain the principal axis direction of the contour. The principal axis refers to the longest axis of the contour, which is usually consistent with the long axis direction of the contour and is used to describe the trend of the contour. This information can be used for subsequent image processing, such as aligning images, measuring body size parameters, etc.

[0131] In one embodiment, within a preset length-width ratio range of the side profile of the mutton sheep, with the centroid of the outer contour as the center, the intercept window is dynamically adjusted in combination with the principal axis direction of the outer contour to align the intercept window with the body trend of the mutton sheep, and the target side profile view of the mutton sheep is intercepted. Exemplarily, first, the interception area is determined. With the centroid as the center, according to the preset length-width ratio range of the mutton sheep side profile (for example, the length is L, the width is W, and the value of L / W is in the interval [a, b]), in combination with the size of the contour, the size of the intercept window is dynamically adjusted. At the same time, the intercept window is rotated with reference to the principal axis direction to align it with the body trend of the mutton sheep, and the side profile view of the target mutton sheep is accurately intercepted to reduce background interference.

[0132] In step S16, at preset angle intervals, the Hough line detection algorithm is used to detect the target side profile view to obtain multiple initial lines, and pruning operations are performed on the multiple initial lines. Then, a spine feature curve is obtained by fitting multiple segmented lines, including:

[0133] The target side profile view is scanned at preset angle intervals to extract the detection points in the image;

[0134] The detection points are mapped to the parameter space, and a vote is taken for each possible line parameter through an accumulator;

[0135] The maximum value of the voting result is found in the accumulator, and the detection points corresponding to the maximum value are connected as the detected initial line;

[0136] The lines in the initial lines with lengths less than the preset length threshold are removed, and the lines in the initial lines with a slope similarity greater than the preset similarity threshold are merged to obtain the pruned lines;

[0137] Multiple lines consistent with the spine trend are selected from the pruned lines to fit the spine feature curve.

[0138] It should be noted that the Hough line detection algorithm is an image processing method used to extract lines from an image. In the target side profile view, the image is scanned at intervals of a preset angle (for example, every 1° or 5°), and the edge points in the image are mapped to the parameter space (polar coordinate space). Multiple initial lines are detected through the accumulator voting mechanism. Specifically, edge detection is performed on the target side profile view (such as using the Canny edge detection algorithm) to extract the edge points in the image. The edge points are transformed into the parameter space, and the accumulator votes on each possible line parameter. Local maxima are searched for in the accumulator, and the parameters corresponding to these maxima are the detected initial lines. Then, pruning operations are carried out. The purpose of the pruning operation is to screen out the line segments related to the spine features from the multiple detected initial lines. The specific steps are as follows: Remove short lines: According to a preset length threshold (such as a line length less than a certain fixed value), too short lines are removed because these lines may belong to noise. Merge approximate lines: Lines with similar slopes and intercepts are merged into one line to reduce duplicate detections. Retain key line segments: According to the anatomical features of the spine, line segments consistent with the spine orientation are selected. For example, line segments located in the middle region of the sheep's back and approximately parallel are selected, and the spine feature curve is obtained by fitting. Multi-segment line fitting can use Bezier curve fitting or spline interpolation, and then the key line segments are extracted as the spine features. Experimental data shows that the average curvature of the spine in the side view of meat sheep is 15°, and the actual curve can be effectively approximated through segmented line fitting (each segment length ≥ 50 pixels), with an error rate ≤ 3%.

[0139] Exemplarily, assume the following initial lines are detected in the target side profile view:

[0140] Line 1: The length is 100 pixels, the slope is 0.1, and the intercept is 50.

[0141] Line 2: The length is 50 pixels, the slope is 0.1, and the intercept is 55.

[0142] Line 3: The length is 80 pixels, the slope is 0.15, and the intercept is 60.

[0143] Line 4: The length is 30 pixels, the slope is 0.2, and the intercept is 70.

[0144] After the pruning operation:

[0145] Remove short straight lines: The length of line 4 is 30 pixels, which is less than the preset length threshold (assumed to be 40 pixels), so it is excluded; Merge approximate straight lines: The slopes and intercepts of line 1 and line 2 are similar, and they can be merged into one straight line. The parameters of the new straight line are slope 0.1 and intercept 52.5; Retain key line segments: The merged line 1 and line 3 have similar slopes and are located in the middle area of the sheep's back, which conforms to the vertebral characteristics, so they are retained as the vertebral characteristic curve. Through the above steps, the finally obtained vertebral characteristic curve can be used for subsequent body size parameter calculation.

[0146] In step S17, determine the endpoint coordinates of the vertebral characteristic curve, and calculate the body size parameters of the meat sheep according to the endpoint coordinates.

[0147] Among them, the relationship between the body size parameters of the meat sheep and the endpoint coordinates is expressed as:

[0148]

[0149] In the formula, S represents the body size parameter of the meat sheep, that is, the estimated body length value of the meat sheep; k represents the proportionality coefficient, m represents the number of vertebral characteristic curves, L i represents the length of the i-th line segment, α i represents the angle between the i-th line segment and the horizontal direction; (x1, y1),..., (x i , y i ), (x i+1 , y i+1 ) are the endpoint coordinates in the vertebral characteristic curve in sequence.

[0150] Among them, in practical applications, a camera calibration process is adopted. For example, a calibration board with a known size (such as a checkerboard) is used for shooting to establish the proportionality coefficient between pixel coordinates and actual lengths, so as to determine the proportionality coefficient. For example, in the system initialization stage, a 20cm×20cm checkerboard calibration board is used for camera calibration, the proportionality coefficient is calculated, and it is applied to the body length estimation formula.

[0151] It should be noted that this formula is used to calculate the body size parameters of meat sheep, that is, the estimated body length of meat sheep. The body length of meat sheep is estimated by calculating the length and direction in the vertebral characteristic curve. This method reasonably utilizes the geometric characteristics of the vertebral column of meat sheep because the vertebral column usually extends along the body length direction of meat sheep. Among them, the proportionality coefficient is used to adjust the calculation result to make it more in line with the actual body length of meat sheep. This coefficient can be calibrated according to experimental data to improve the accuracy of the calculation. By considering multiple vertebral characteristic curves, the body length of meat sheep can be more comprehensively reflected. This method is more accurate than estimating with only a single line segment. At the same time, by considering the length of each line segment and the angle with the horizontal direction, the actual trend of the body length of meat sheep can be more accurately reflected because the vertebral column of meat sheep is not completely horizontal. By comprehensively considering the length and direction of multiple vertebral characteristic curves, the formula can provide a more accurate estimated value of the body length. This method is more reliable than simple straight-line measurement. In summary, this embodiment is used in combination with a machine vision system to achieve the automatic measurement of the body size parameters of meat sheep. This is very useful for large-scale farms because it can reduce the workload and errors of manual measurement.

[0152] In summary, the present invention provides an efficient and accurate automatic measurement solution through a series of image processing technologies. First, the original side image of the meat sheep is obtained and standardized to obtain a standard meat sheep image, avoiding errors caused by differences in shooting angles and image quality. Then, the epidermis area of the meat sheep is extracted using an image clustering algorithm, and the epidermis pixels are dilated to further optimize the image quality and ensure the integrity of the shape features. Then, the standard meat sheep image is intercepted through the epidermis mask to obtain a side sectional view of the target meat sheep, providing key data for subsequent body size measurement. On this basis, a straight-line detection algorithm is applied to identify the body shape feature lines of the meat sheep, and the vertebral characteristic curve is obtained through effective straight-line integration. Furthermore, the body length parameter of the meat sheep is accurately calculated by combining the endpoint coordinates of the vertebral characteristic curve. The present invention avoids the errors of traditional manual measurement or relying on simple algorithms, improves the accuracy of meat sheep body size measurement, and at the same time can improve the measurement efficiency compared with manual measurement.

[0153] Referring to Figure 2 , the second embodiment of the present invention provides a machine vision-based meat sheep body size measurement system, including:

[0154] An image acquisition module for acquiring the original side image of the meat sheep;

[0155] An image standardization module for standardizing the original side image according to a preset aspect ratio to obtain a standard meat sheep image;

[0156] A clustering module, configured to calculate a silhouette coefficient according to the standard mutton sheep images, determine the number of clustering centers according to the silhouette coefficient, and perform clustering based on the standard mutton sheep images and the number of clustering centers to obtain clustering centers;

[0157] A mask acquisition module, configured to select an epidermal pixel range according to the clustering centers to obtain epidermal pixels, and perform a dilation operation on the epidermal pixels to obtain an epidermal mask;

[0158] An image cropping module, configured to extract a contour of the epidermal mask, and use a contour detection algorithm to identify the centroid of the outer contour and the direction of the major axis of the outer contour of the epidermal region; within a preset length-width ratio range of the side profile of the mutton sheep, with the centroid of the outer contour as the center, dynamically adjust a cropping window in combination with the direction of the major axis of the outer contour to align the cropping window with the body orientation of the mutton sheep, and crop to obtain a target side profile view of the mutton sheep;

[0159] A straight line detection module, configured to detect a plurality of initial straight lines from the target side profile view at preset angle intervals by using a Hough straight line detection algorithm, and perform a pruning operation on the plurality of initial straight lines to obtain a spine feature curve;

[0160] A body measurement calculation module, configured to determine the endpoint coordinates of the spine feature curve, and calculate mutton sheep body measurement parameters according to the endpoint coordinates.

[0161] It should be noted that a mutton sheep body measurement system based on machine vision provided in an embodiment of the present invention is used to execute all process steps of a mutton sheep body measurement method based on machine vision in the above embodiment. The working principles and beneficial effects of the two correspond one by one, and thus will not be elaborated herein.

[0162] An embodiment of the present invention further provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as an image acquisition program. When the processor executes the computer program, the steps in the above embodiments of each mutton sheep body measurement method based on machine vision are implemented, such as Figure 1 step S11 shown. Alternatively, when the processor executes the computer program, the functions of each module / unit in the above device embodiments are implemented, such as the image acquisition module.

[0163] Exemplarily, the computer program may be divided into one or more modules / units. The one or more modules / units are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device.

[0164] The electronic device can be a computing device such as a desktop computer, notebook, palm computer, and smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the above components are only examples of the electronic device and do not constitute a limitation on the electronic device. It may include more or fewer components than the above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0165] The so-called processor can be a Central Processing Unit (CPU), or other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The processor is the control center of the electronic device, connecting various parts of the entire electronic device through various interfaces and lines.

[0166] The memory can be used to store the computer program and / or module. The processor realizes various functions of the electronic device by running or executing the computer program and / or module stored in the memory, and by calling the data stored in the memory. The memory mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function (such as a sound playback function, an image playback function, etc.); the data storage area can store data created according to the use of the mobile phone (such as audio data, phone book, etc.). In addition, the memory can include high-speed random access memory, and can also include non-volatile memory, such as a hard disk, memory, plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, Flash Card, at least one magnetic disk storage device, flash device, or other volatile solid-state storage devices.

[0167] Among them, if the modules / units integrated in the electronic device are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, to implement all or part of the processes in the above-described embodiment methods of the present invention, it can also be completed by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-described various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0168] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which can be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0169] The above-described specific embodiments have further elaborated on the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only the specific embodiments of the present invention and is not used to limit the protection scope of the present invention. It is particularly pointed out that for those skilled in the art, any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for measuring the body size of meat sheep based on machine vision, characterized in that, Including: Obtain the original side image of the meat sheep; Standardize the original side image according to a preset length-width ratio to obtain a standard meat sheep image; Calculate the silhouette coefficient based on the standard meat sheep image, determine the number of cluster centers according to the silhouette coefficient, and perform clustering based on the standard meat sheep image and the number of cluster centers to obtain cluster centers; Select the epidermal pixel range according to the cluster centers to obtain epidermal pixels, and perform a dilation operation on the epidermal pixels to obtain an epidermal mask; Extract the contour of the epidermal mask, and use a contour detection algorithm to identify the centroid of the outer contour and the direction of the main axis of the outer contour in the epidermal region; within the preset length-width ratio range of the side profile of the meat sheep, with the centroid of the outer contour as the center, dynamically adjust the cropping window in combination with the direction of the main axis of the outer contour to align the cropping window with the body orientation of the meat sheep, and crop to obtain the target side profile view of the meat sheep; At preset angle intervals, use the Hough line detection algorithm to detect the target side profile view to obtain multiple initial lines, and perform pruning operations on the multiple initial lines, and then fit the multiple segmented lines to obtain a spine feature curve; Determine the endpoint coordinates of the spine feature curve, and calculate the body size parameters of the meat sheep according to the endpoint coordinates.

2. The method for measuring the body size of meat sheep based on machine vision according to claim 1, wherein Obtain the original side image of the meat sheep, including: Use a collection device to obtain the side image of the meat sheep, and send the image data to the data center of the system server; According to the side image of the meat sheep, judge whether the image resolution is greater than the preset resolution. If not, discard the image; if so, use bicubic interpolation to adjust the image resolution to the preset resolution to obtain the initial side image of the meat sheep; According to the initial side image of the meat sheep, judge whether the image file size is greater than the preset size threshold. If so, perform image compression; if not, directly obtain the original side image of the meat sheep.

3. The method for measuring the body size of meat sheep based on machine vision according to claim 1, characterized in that Calculate the silhouette coefficient based on the standard meat sheep image, determine the number of cluster centers according to the silhouette coefficient, and perform clustering based on the standard meat sheep image and the number of cluster centers to obtain cluster centers, including: Calculate the silhouette coefficient according to the standard meat sheep image through the following formula: where a(i) represents the average distance from sample i to other samples in the same cluster, b(i) represents the minimum distance from sample i to other samples in the same cluster, and s(i) is the silhouette coefficient; Traverse different K values, calculate the average value of the silhouette coefficients of all samples, and select the K value that maximizes the average value of the silhouette coefficients as the number of cluster centers; Perform clustering on the standard meat sheep image according to the number of cluster centers to obtain cluster centers.

4. The method for measuring the body size of meat sheep based on machine vision according to claim 1, wherein Select the epidermal pixel range according to the cluster centers to obtain epidermal pixels, and perform a dilation operation on the epidermal pixels to obtain an epidermal mask, including: According to the cluster centers, calculate the pixel point density with each cluster center as the core; Select the epidermal pixel range according to the pixel point density to obtain epidermal pixels; Perform a dilation operation on the epidermal pixels according to the cluster centers to obtain an epidermal mask.

5. The method for measuring the body size of meat sheep based on machine vision according to claim 4, characterized in that, The performing a dilation operation on the epidermal pixels according to the cluster centers to obtain an epidermal mask includes: Divide the preset small areas according to the epidermal pixels, and calculate the regional entropy value; According to the regional entropy value and the small area, determine whether each small area is higher than a preset entropy value threshold, if so, the skin color range of the area is enlarged, if not, no operation is performed, and the optimized epidermal pixels are obtained; According to the optimized epidermal pixels, a morphological dilation operation is applied to connect scattered epidermal pixel points into a complete area to obtain an initial epidermal mask; The initial epidermis mask is preprocessed to obtain an epidermis mask.

6. The method for measuring the body dimensions of meat sheep based on machine vision according to claim 5, wherein The preprocessing of the initial epidermis mask to obtain the epidermis mask comprises: Removing patches caused by uneven brightness in the initial epidermal mask to obtain a uniform epidermal mask; De-noising the uniform epidermis mask by using a median filter to obtain a de-noised epidermis mask; The denoised epidermal mask is normalized to obtain an epidermal mask.

7. The method for measuring the body size of meat sheep based on machine vision according to claim 1, wherein The step of extracting the contour of the epidermis mask and identifying the outer contour centroid and the outer contour main axis direction of the epidermis region by using a contour detection algorithm includes: Converting the epidermis mask into a grayscale image and performing binarization processing to obtain a binary image; Scan the binary image using a contour detection algorithm to find a closed boundary of the epidermal area and form a polygonal contour; Calculate the geometric center of the polygonal contour by image moment to obtain the centroid of the outer contour; The polygonal contour is fitted using a minimum area rectangle, and the long side direction of the minimum area rectangle is used as the main axis direction of the outer contour.

8. The method for measuring the body dimensions of meat sheep based on machine vision according to claim 1, characterized in that, The method of detecting the target side profile view using the Hough line detection algorithm at intervals of preset angles to obtain a plurality of initial lines, and performing pruning operations on the plurality of initial lines, and then obtaining a spine characteristic curve by fitting a plurality of segmented lines, includes: Scanning the target side profile view at intervals of a preset angle to extract detection points in the image; Mapping the detection points to the parameter space, and voting for each possible line parameter through an accumulator; Find the maximum value of the voting result in the accumulator, and connect the detection points corresponding to the maximum value as the detected initial straight line; Eliminate the straight lines whose lengths are less than a preset length threshold in the initial straight lines, and merge the straight lines whose slopes are similar to each other and greater than a preset similarity threshold in the initial straight lines to obtain pruned straight lines; Multiple straight lines that are consistent with the spine direction are selected from the pruned straight lines, and the spine characteristic curve is obtained by fitting.

9. The method for measuring the body size of meat sheep based on machine vision according to claim 1, wherein, The step of calculating the body size parameters of the mutton sheep according to the endpoint coordinates includes: The relationship between the mutton body size parameters and the endpoint coordinates is expressed as: In the formula, S represents the body size parameter of the meat sheep, that is, the estimated body length of the meat sheep; k represents the proportionality coefficient, m represents the number of vertebral characteristic curves, and L i represents the length of the i-th line segment, and α i represents the angle between the i-th line segment and the horizontal direction; (x1, y1),..., (x i , y i ), (x i+1 , y i+1 ) are the endpoint coordinates of the line segments in the vertebral characteristic curve in sequence.

10. A mutton sheep body measurement system based on machine vision, characterized in that, include: An image acquisition module, used to acquire an original side image of the sheep; An image standardization module is used to perform standardization processing on the original side image according to a preset aspect ratio to obtain a standard mutton sheep image; A clustering module, used for calculating a silhouette coefficient according to the standard mutton image, determining the number of cluster centers according to the silhouette coefficient, and performing clustering according to the standard mutton image and the number of cluster centers to obtain cluster centers; A mask acquisition module, used for selecting an epidermal pixel range according to the cluster center to obtain epidermal pixels, and performing a dilation operation on the epidermal pixels to obtain an epidermal mask; An image cropping module, configured to extract the contour of the epidermal mask, and use a contour detection algorithm to identify the centroid of the outer contour and the direction of the major axis of the outer contour of the epidermal region; within a preset length-width ratio range of the side profile of the mutton sheep, centered on the centroid of the outer contour, dynamically adjust the cropping window in combination with the direction of the major axis of the outer contour, align the cropping window with the body orientation of the mutton sheep, and crop to obtain a target side profile view of the mutton sheep; A straight line detection module, configured to detect a plurality of initial straight lines from the target side profile view at preset angle intervals by using a Hough straight line detection algorithm, and perform pruning operations on the plurality of initial straight lines to obtain a spine feature curve; A body measurement calculation module, configured to determine the endpoint coordinates of the spine feature curve, and calculate mutton sheep body measurement parameters based on the endpoint coordinates.

Images