A method and system for controlling a meat processing process

Abstract

This invention provides a method and system for controlling the processing of meat products, belonging to the field of meat product processing. First, the surface image data of raw meat is enhanced and preprocessed to obtain stable feature image data of the raw meat. Based on the stable feature image data, enhanced texture image data of the raw meat is obtained, and the distribution boundary of the raw meat components is determined. The distribution boundary of the raw meat components is proportionally extracted to obtain the proportion data of the raw meat components. Based on the proportion data of the raw meat components, low fascia location data is obtained. Based on the low fascia location data, an initial raw meat cutting path is determined, and an initial cutting path is obtained. A path adjustment strategy is executed on the initial cutting path to obtain the final raw meat cutting path. Based on the final raw meat cutting path, a final cutting command is obtained, and the cutting module is driven to perform the cutting operation according to the final cutting command. This invention provides a meat product processing control method that can enhance color recognition and ensure accurate separation of raw meat based on color recognition.

Description

Technical Field

[0001] This invention relates to the field of meat processing technology, and more specifically, to a method and system for controlling the meat processing process. Background Technology

[0002] Meat processing is an important part of the food industry. Its core lies in how to process meat products efficiently and precisely to ensure the quality and production efficiency of meat products. With the increasing demand for meat products from consumers, accurately identifying and controlling the proportion of lean meat, fat, and fascia in meat products has become the key to improving the cutting technology of processed meat products. However, current traditional meat processing methods often struggle to achieve more refined control when faced with the relatively complex structure of meat.

[0003] In traditional meat processing, the tissue composition of meat products is often identified by manual inspection or simple machine sorting. Manual inspection relies on the experience of the staff, which is inefficient and prone to errors. Especially when facing large-scale production, prolonged inspection of meat products can cause eye strain and other problems, making it impossible to guarantee consistency. While mechanical sorting using traditional machine vision can improve efficiency, it is usually based on a single criterion, which affects the subsequent accurate cutting of meat products.

[0004] On the one hand, although lean meat, fat, and fascia differ in color, these differences can become blurred under different lighting conditions or meat freshness. For example, on a factory assembly line, changes in light or differences in surface moisture of the meat can make it difficult to extract color features stably, thus affecting the accurate calculation of component ratios. For instance, in actual processing scenarios, a piece of pork may have lean meat that appears similar in color to fat meat due to different lighting angles, making it difficult for traditional machine vision to extract color features stably, thereby affecting the calculation of component ratios. On the other hand, even after the color features are stably extracted and the component ratios are obtained, how to accurately and completely separate lean meat, fat, and fascia based on the component ratios remains a common problem.

[0005] In summary, this invention provides a meat processing control method that enhances color recognition and ensures precise separation of raw meat during the cutting process based on color recognition. Summary of the Invention

[0006] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a method for controlling the processing of meat products, the method comprising:

[0007] Obtain raw meat surface image data, perform enhancement preprocessing on the raw meat surface image data, and obtain raw meat stable feature image data;

[0008] Based on the original meat stable feature image data, obtain the original meat texture enhancement image data, determine the original meat component distribution boundary, extract the proportion of the original meat component distribution boundary, obtain the original meat component proportion data, and obtain the low fascia location data based on the original meat component proportion data.

[0009] An initial raw meat cutting path is obtained based on low fascia location data. A path adjustment strategy is then applied to the initial cutting path to obtain the final raw meat cutting path.

[0010] The final cutting instruction is obtained based on the final cutting path of the raw meat, and the cutting module is driven to perform the cutting operation according to the final cutting instruction.

[0011] As a further aspect of the present invention, obtaining raw meat surface image data, performing enhanced preprocessing on the raw meat surface image data, and obtaining raw meat stable feature image data includes:

[0012] The raw meat is photographed in real time using a camera to obtain surface image data of the raw meat. Based on the surface image data of the raw meat, pixel histogram data of the raw meat is obtained, and pixel distribution data of the raw meat is obtained based on the pixel histogram data of the raw meat.

[0013] The original meat pixel distribution data is balanced and adjusted using a gray-level cumulative distribution function to obtain balanced image data of the original meat. The gray-level range is determined based on the balanced image data of the original meat, and the extended image data of the original meat is obtained through the gray-level range.

[0014] Color analysis is performed on the extended image data of the raw meat to extract the color distribution data of the raw meat. Stability analysis is then performed on the color distribution data of the raw meat to obtain stable feature image data of the raw meat.

[0015] As a further aspect of the present invention, based on the original meat stable feature image data, original meat texture enhancement image data is obtained, and the original meat component distribution boundary is determined. The original meat component distribution boundary is then proportionally extracted to obtain original meat component proportion data. Based on the original meat component proportion data, low fascia location data is obtained, including:

[0016] The brightness of the original meat stable feature image is adjusted based on gamma correction to obtain the original meat contrast enhancement image data. At the same time, the texture of the original meat contrast enhancement image is enhanced to obtain the original meat texture enhancement image data.

[0017] The boundary of the original meat component distribution boundary is obtained by performing boundary extraction on the original meat texture enhancement image data using a component extraction strategy.

[0018] Based on the distribution boundary of the raw meat components, the total boundary pixel data is obtained, and the proportion data of the raw meat components is obtained based on the total boundary pixel data. At the same time, the proportion data of the raw meat components is updated in real time based on the update mechanism.

[0019] Based on the proportion of fascia components in the original meat composition data, a proportion judgment is made to determine the low fascia location data.

[0020] As a further aspect of the present invention, a component extraction strategy is used to extract the boundaries of the original meat texture enhancement image to obtain the distribution boundaries of the original meat components, including:

[0021] The lean meat component region in the original meat texture enhancement image data is segmented based on a clustering algorithm to obtain lean meat component region data;

[0022] The fat component region in the original meat texture enhancement image data is segmented based on threshold segmentation to obtain fat component region data;

[0023] The fascia component region in the original meat texture enhancement image data is segmented based on the Canny edge detection algorithm to obtain fascia component region data;

[0024] Contour detection is performed on the lean meat component region data, the fat meat component region data, and the fascia component region data to obtain the fat-lean-tendon boundary data, and the original meat component distribution boundary is obtained based on the fat-lean-tendon boundary data.

[0025] As a further aspect of the present invention, the total boundary pixel data is obtained based on the distribution boundary of the original meat components, the proportion data of the original meat components is obtained based on the total boundary pixel data, and the proportion data of the original meat components is updated in real time based on an update mechanism, including:

[0026] The boundary of the original meat component distribution is processed based on the contour tracking algorithm to obtain the number of pixels of lean meat boundary, the number of pixels of fat boundary, and the number of pixels of fascia boundary.

[0027] The total number of boundary pixels is obtained based on the number of pixels at the boundaries of lean meat, fat, and fascia.

[0028] The original meat component ratio data is obtained by calculating the total boundary pixel data, which includes lean meat component ratio data, fat meat component ratio data and fascia component ratio data.

[0029] When acquiring the raw meat component ratio data, an update mechanism is used to obtain the raw meat component ratio data deviation based on the input raw meat texture enhancement image data, and the raw meat component ratio data is updated in real time based on the raw meat component ratio data deviation.

[0030] As a further aspect of the present invention, a proportion determination is made based on the proportion data of fascia components in the original meat component proportion data to determine the low fascia location data, including:

[0031] If the proportion of the fascia components is greater than or equal to a predefined fascia threshold, then the pixel density data of the fascia region is obtained based on the original meat component distribution boundary, and the range of the low fascia region is determined based on the pixel density data of the fascia region.

[0032] Based on the range of the low fascia region, obtain the feature point set data of the low fascia region, group the feature point set data of the low fascia region, obtain the pixel distribution data of the region range, and confirm the center data of the low fascia location based on the pixel distribution data of the region range.

[0033] The low fascia location center data is optimized to obtain optimized low fascia location center data. The mean pixel data and variance data of the region are calculated based on the pixel density data of the fascia region. The low fascia location data is obtained through the optimized low fascia location center data, the mean pixel data of the region, and the variance data of the region.

[0034] If the proportion of fascia components is less than a predefined fascia threshold, then low fascia location data are obtained based on the original meat component distribution boundary.

[0035] As a further aspect of the present invention, an initial raw meat cutting path is obtained based on low fascia location data, and a path adjustment strategy is applied to the initial cutting path to obtain the final raw meat cutting path, including:

[0036] The initial raw meat cutting path is generated based on the A* search algorithm;

[0037] The A-star search algorithm takes low fascia location data as input and uses the low fascia location data as the starting point to perform path search on the original meat texture enhancement image data.

[0038] Obtain raw meat component distribution data, and perform a path adjustment strategy on the initial raw meat cutting path based on the raw meat component distribution data to obtain the final raw meat cutting path.

[0039] As a further aspect of the present invention, obtaining raw meat component distribution data, and performing a path adjustment strategy on the initial raw meat cutting path based on the raw meat component distribution data to obtain the final raw meat cutting path, including:

[0040] Based on the enhanced image of the original meat texture and the boundary of the original meat component distribution, the original meat component distribution data is obtained, and at the same time, the high fascia location data is obtained based on the low fascia location data.

[0041] The fat separation area is determined based on the original meat component distribution data, and obstacle coordinate data is obtained based on the high fascia location data and the obstacle acquisition mechanism.

[0042] The initial raw meat cutting path is corrected and adjusted by using the obstacle coordinate data and the fat separation area to determine the final raw meat cutting path, i.e., the final raw meat cutting path.

[0043] Furthermore, embodiments of the present invention also provide a meat product processing control system, the system comprising:

[0044] The enhancement module is used to acquire raw meat surface image data and raw meat stability feature image data;

[0045] A boundary confirmation module is used to acquire enhanced image data of the original meat texture and determine the distribution boundary of the original meat components;

[0046] The location confirmation module is used to perform proportion extraction to obtain raw meat component proportion data and low fascia location data.

[0047] The instruction generation module is used to obtain an initial cutting path, execute a path adjustment strategy to obtain the final cutting path of the raw meat, and obtain the final cutting instruction based on the final cutting path of the raw meat.

[0048] A cutting module, which is used to cut the raw meat;

[0049] A driving module is used to drive the cutting module to perform a cutting operation according to the final cutting command.

[0050] Based on the above, this application embodiment first acquires raw meat surface image data, performs enhanced preprocessing on the raw meat surface image data to acquire stable feature image data of raw meat, and eliminates interference information in the raw meat surface image, such as reflection, uneven brightness, etc. Then, based on the stable feature image data of raw meat, it acquires raw meat texture enhanced image data and determines the raw meat component distribution boundary. The raw meat component distribution boundary is proportionally extracted to acquire raw meat component proportion data. Based on the raw meat component proportion data, it acquires low fascia position data to provide coordinate basis for subsequent cutting path planning. Then, based on the low fascia position data, it performs an initial raw meat cutting path and acquires the initial cutting path. The initial cutting path is then adjusted using a path adjustment strategy to acquire the final raw meat cutting path. Finally, based on the final raw meat cutting path, it acquires the final cutting command and drives the cutting module to perform the cutting operation according to the final cutting command. Through multi-level image analysis and dynamic path planning, it achieves accurate separation of raw meat and maximizes the utilization rate of raw meat. It provides a meat product processing control method that can enhance color recognition and ensure accurate separation of raw meat by the cutting process based on color recognition. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the execution flow of a meat product processing control method provided in an embodiment of the present invention.

[0052] Figure 2 This is a schematic diagram of the execution flow of step S2 in a meat product processing control method provided in an embodiment of the present invention.

[0053] Figure 3 This is a schematic diagram of a meat product processing control system provided in an embodiment of the present invention. Detailed Implementation

[0054] The accompanying drawings in the embodiments provide a clear and complete description of the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can typically be arranged and designed in various different configurations.

[0055] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0056] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a schematic diagram of the execution flow of a meat product processing control method according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the execution flow of step S2 in a meat product processing control method provided in an embodiment of the present invention. The following is a detailed description of this meat product processing control method.

[0057] Specifically, a method for controlling the processing of meat products includes:

[0058] Specifically, a method for processing electricity metering and detection data with multi-scenario indicator settings includes:

[0059] Step S1: Obtain raw meat surface image data, perform enhancement preprocessing on the raw meat surface image data, and obtain raw meat stable feature image data.

[0060] In this embodiment, step S1 includes:

[0061] Step S11: Take real-time photos of the raw meat using a camera to obtain surface image data of the raw meat, obtain pixel histogram data of the raw meat based on the surface image data of the raw meat, and obtain pixel distribution data of the raw meat based on the pixel histogram data of the raw meat.

[0062] Specifically, the raw meat is photographed in real time in a standardized environment using a pre-deployed high-resolution camera to avoid image differences caused by environmental changes, thus obtaining surface image data of the raw meat. Then, the surface image data of the raw meat can be calculated using the OpenCV library. During the calculation, the surface image data of the raw meat is first converted into grayscale format, and then the number of pixels at each grayscale level is obtained, thereby obtaining the pixel histogram of the surface image data of the raw meat, that is, the raw meat pixel histogram data. In the raw meat pixel histogram, the horizontal axis is the grayscale value, and the vertical axis is the number of pixels at the corresponding grayscale. Then, by counting the number of pixels at each grayscale level, the pixel distribution data of the raw meat is obtained.

[0063] Understandably, a standardized environment means shooting the raw meat in a standardized and fixed environment with a fixed shooting angle and distance. The selected high-resolution camera can be a 1920*1080 high-resolution camera with a frame rate of 30 or 60fps to ensure that the captured raw meat surface image data is clear.

[0064] In this embodiment, taking raw beef as an example, an image of raw beef can be captured at a height of 30cm above the surface, at a vertical angle. The camera captures the surface of the raw beef at high resolution. The surface of the raw beef includes details such as color, texture, and lighting, which can directly reflect the freshness of the meat product and potential defects such as discoloration or contamination. This provides the most basic data for subsequent processing steps. After acquisition, OpenCV is used to calculate and generate a pixel histogram of the raw beef surface image. Then, by counting the number of pixels at each gray level in the raw beef pixel histogram, the pixel distribution data of the raw beef is generated. If the pixel distribution data of the raw beef image shows that most pixels are concentrated in the low to medium gray values, it may indicate that the image is relatively dark. This pixel distribution data can quantify the overall brightness distribution of the image, preparing for subsequent equalization adjustments and preventing uneven lighting problems.

[0065] Step S12: The pixel distribution data of the original meat is balanced and adjusted using a gray-level cumulative distribution function to obtain balanced image data of the original meat. The gray-level range is determined based on the balanced image data of the original meat, and the extended image data of the original meat is obtained through the gray-level range.

[0066] Specifically, the original meat pixel distribution data is histogram equalized using the cumulative distribution function (CDF) to stretch the gray range in the original meat pixel distribution data to a wider range, making the overall gray distribution of the image more uniform, thereby improving the contrast of different component regions. Then, based on the equalized gray level, the effective gray range is determined and the gray range is expanded to avoid the loss of edge information of the image. Finally, the original meat extended image data can be obtained.

[0067] Understandably, the effective grayscale range refers to the grayscale range excluding the background.

[0068] For example, after obtaining the pixel distribution data of raw beef, suppose we find that the grayscale of the fat is concentrated in the range of 200-220 and that of the lean meat is 80-120. The fat may be overexposed, causing blurry details, while the lean meat may be too dark, resulting in unclear texture. We can then use the cumulative grayscale distribution function to balance the pixel distribution data. Suppose we expand the fat range to 150-220 and the lean meat to 100-140, the overall contrast of the image will be improved, making the meat texture clearer, and we will obtain a balanced image of the raw beef. Then we can further analyze the balanced image. For example, we can use OpenCV to find the minimum and maximum grayscale values ​​to determine the grayscale range and remove invalid grayscale intervals to determine the effective grayscale range. We can then linearly expand the effective grayscale range by stretching the range according to the formula: pixel value = (old value - minimum value) / (maximum value - minimum value) * 255, thereby generating expanded image data of the raw beef.

[0069] Step S13: Perform color analysis on the extended image data of the raw meat, extract the color distribution data of the raw meat, perform stability analysis on the color distribution data of the raw meat, and obtain stable feature image data of the raw meat.

[0070] Specifically, the RGB three-channel color analysis is performed on the raw meat extended image data to extract the color distribution data of each component, thereby establishing the correspondence between color and raw meat. Then, the color distribution data of raw meat can be stabilized. Stable color clusters in the distribution are selected by thresholding and long-term stable color clusters are retained to ensure that the final obtained stable feature image data of raw meat is stable data and to prevent abnormal color areas from interfering with subsequent steps.

[0071] Understandably, step S1 allows for the acquisition and enhancement of basic image data, eliminating interference information in the original meat surface image, such as reflection and uneven brightness, and extracting image data that stably reflects the image characteristics. This provides a data foundation for component analysis and cutting path planning in subsequent steps. Through the progressive process of image acquisition, grayscale equalization, and color stability analysis, the conversion from basic image data to stable image data is achieved.

[0072] Step S2: Based on the original meat stable feature image data, obtain the original meat texture enhancement image data, determine the original meat component distribution boundary, extract the proportion of the original meat component distribution boundary, obtain the original meat component proportion data, and obtain the low fascia location data based on the original meat component proportion data.

[0073] In this embodiment, step S2 includes:

[0074] Step S21: Adjust the brightness of the raw meat stable feature image based on gamma correction to obtain raw meat contrast enhancement image data, and simultaneously enhance the texture of the raw meat contrast enhancement image to obtain raw meat texture enhancement image data.

[0075] Specifically, after acquiring the stable feature image data of the raw meat, the brightness of the stable feature image data is adjusted so that the brightness of the image can better highlight the texture details of the image. The brightness of the stable feature image data of the raw meat is adjusted by gamma correction to obtain the contrast enhancement image data of the raw meat. At the same time, the texture of the contrast enhancement image data of the raw meat can be enhanced by combining Gaussian filtering and Laplacian operator. Gaussian filtering can smooth image noise, while Laplacian operator can highlight image edges and textures, such as the fiber texture of fascia and the muscle texture of lean meat. Finally, the texture enhancement image data of the raw meat is obtained, and the accuracy of subsequent component boundary extraction is ensured based on the texture enhancement image data of the raw meat.

[0076] As is understandable, gamma correction adjusts brightness using the formula: Output pixel value = Input pixel value^γ, where γ represents the gamma power and is used to describe the non-linear relationship between image brightness and pixel value. By adjusting the γ value, the brightness and contrast of the image can be changed.

[0077] Furthermore, the gamma power in gamma correction is dynamically adjusted based on the real-time brightness of the image. When the image is too dark, the gamma power is less than 1 to increase the brightness, while when the image is too bright, the gamma power is greater than 1 to decrease the brightness.

[0078] In some possible embodiments, taking raw beef meat as an example, if the brightness of the stable feature image data of raw beef meat is too dark, the brightness of the image can be adjusted by γ=0.8 to obtain the contrast-enhanced image data of raw beef meat. Then, the contrast-enhanced image data of raw beef meat is subjected to Gaussian filtering with a standard deviation of 1.2 to remove small noise in the image. Then, the image gradient is calculated by the 3*3 Laplacian algorithm to highlight the continuous fiber texture of the fascia and transform the originally blurred fascia edges into clear lines, thereby obtaining the texture-enhanced image data of raw beef meat.

[0079] Step S22: The original meat texture enhancement image data is subjected to boundary extraction using a component extraction strategy to obtain the distribution boundary of the original meat components.

[0080] In this embodiment, step S22 includes:

[0081] Step S221: Segment the lean meat component region in the original meat texture enhancement image data based on a clustering algorithm to obtain lean meat component region data.

[0082] Specifically, a clustering algorithm such as K-means is used for region segmentation. The RGB and gray values ​​in the original meat texture enhancement image are used as input to the clustering algorithm, and the number of clusters is set to 3, corresponding to the three components of fat, lean meat and fascia. The clustering algorithm uses an iterative algorithm to group regions with similar RGB and gray values ​​into the same category, and outputs the region of lean meat component through the clustering algorithm.

[0083] Understandably, the image features of lean meat regions are relatively uniform, with subtle, continuous variations and no single, clear segmentation threshold. For example, the tenderloin region of raw beef is generally bright red, but due to the influence of muscle texture, the pixel values ​​at different locations will fluctuate continuously within this range. For instance, the lean meat near the fascia is darker, while the central area is brighter, making it impossible to completely separate it from other regions using a fixed threshold. However, the K-means clustering algorithm is an algorithm that does not require a preset threshold and can automatically group pixels based on the similarity of their features, such as RGB and grayscale values. Even if the pixel values ​​of the lean meat region fluctuate continuously, as long as the overall features are consistent, it can be clustered, thus segmenting the complete lean meat region. If a threshold segmentation is used to process the lean meat region, some brighter or darker lean meat regions may be misclassified as other regions due to the fixed threshold. On the other hand, if an edge detection algorithm is used, a large number of interfering edges may be extracted due to the complex texture inside the lean meat, making it impossible to generate a complete region.

[0084] For example, when enhancing the texture of a raw beef image, the dataset consisting of the RGB and grayscale values ​​of each pixel is input into the K-means clustering algorithm. The number of clusters k=3, the number of iterations is 50, and the error threshold x is set. After the K-means clustering algorithm iterates, three cluster centers are obtained. Assuming that the cluster centers are center 1, center 2, and center 3, if center 1 matches the color and grayscale characteristics of lean meat, then all pixel regions belonging to center 1 are located as lean meat components.

[0085] Step S222: Segment the fat component region in the original meat texture enhancement image data based on threshold segmentation to obtain fat component region data.

[0086] Specifically, for the image features of fatty meat regions, which are characterized by high brightness, significantly higher grayscale values ​​than other components, and a clearly defined feature threshold, the Otsu adaptive threshold segmentation algorithm can be used. This algorithm does not require manual threshold setting; it automatically finds the optimal threshold by calculating the inter-class variance of the image's grayscale histogram. Then, it locates the fatty meat regions in the original meat texture enhanced image where the grayscale value is greater than or equal to the optimal threshold and the color matches the characteristics of fatty meat, thereby obtaining the fatty meat component region data.

[0087] Understandably, the Otsu adaptive threshold segmentation algorithm can quickly segment fatty regions with high brightness, significantly higher grayscale values ​​than other components, clear feature boundaries, and obvious segmentation thresholds. It can also avoid over-segmentation or under-segmentation. If clustering algorithms are used to process fatty regions, it will lead to a loss of computing power due to the relatively simple features of the fatty regions, and the segmentation effect is almost the same as threshold segmentation. If edge detection algorithms are used, there may be cases where only the edges of the fatty regions are extracted, and the complete fatty regions cannot be obtained directly, requiring additional region filling steps.

[0088] For example, when enhancing the texture of a raw beef image, assuming that the optimal segmentation threshold is calculated to be 160 after analyzing the grayscale histogram using the Otsu adaptive threshold segmentation algorithm, the pixel regions with grayscale values ​​greater than or equal to 160 in the image are extracted. Then, the regions containing fat are selected by combining the RGB features of fat (R220-240, G210-230, B200-220).

[0089] Step S223: Based on the Canny edge detection algorithm, the fascia component region in the original meat texture enhancement image data is segmented to obtain fascia component region data.

[0090] Specifically, the image features of the fascia region are strip-shaped with relatively uniform internal grayscale, but with clear boundaries from the surrounding area, forming a continuous edge contour. The Canny edge detection algorithm is used to segment the fascia components in the original flesh texture enhancement image data. This algorithm extracts the continuous edges of the fascia region by calculating the gradient magnitude and direction, refining the edges, and determining the effective edges. Then, the detected edges are closed by filling the edge gaps to form a complete fascia region contour. The area within the contour that matches the color and grayscale characteristics of the fascia is determined as the fascia region, thereby obtaining the fascia component region data.

[0091] Understandably, the Canny edge detection algorithm can accurately extract continuous edges in an image, obtaining clear and complete edge contours. However, for components like fascia that are narrow but have prominent edges, edge detection can directly capture the contour direction and determine the relevant fascia area by closing the contour, compared to clustering or thresholding. If clustering is used to process the fascia area, it will be covered by most pixels of lean or fat meat because the area of ​​the fascia is too small and the number of pixels is small, causing the cluster center to be biased towards the main part. If thresholding is used, it will be difficult to find a single threshold for accurate segmentation because the gray range of the fascia overlaps with some parts, which can easily lead to misjudgment.

[0092] For example, when enhancing the texture of a raw beef image, the Canny edge detection algorithm can be applied. First, Gaussian filtering with a standard deviation of 0.8 can be used to remove noise. Then, the gradients in the X and Y directions can be calculated using the Sobel operator to obtain the gradient magnitude and direction. Next, non-maximum suppression is performed to refine wide edges into single-pixel edges. Finally, a dual threshold can be set, such as a low threshold of 50 and a high threshold of 150, to retain strong edges with the high threshold and weak edges that connect strong edges with those above the low threshold, thus obtaining continuous edges of the fascia. After the edges are closed, regions whose color and grayscale inside the edges match the characteristics of the fascia are filtered to obtain the regions containing the fascia components.

[0093] Step S224: Perform contour detection on the lean meat component region data, the fat meat component region data, and the fascia component region data to obtain the lean-fat-tendon boundary data, and obtain the original meat component distribution boundary based on the lean-fat-tendon boundary data.

[0094] Specifically, contour detection is performed on the component regions of lean meat, fat meat, and fascia to obtain the contour pixel coordinates of each region. Then, the contour pixel coordinates of each region can be integrated to form a boundary covering all raw meat components, generating fat-lean-fascia boundary data, which reflects the spatial positional relationship of each component of the raw meat.

[0095] Furthermore, contour detection can be performed using functions such as OpenCV's findContours. The findContours function has an external contour search mode and a polygon approximation method, and then contour detection can be performed using the findContours function.

[0096] For example, when extracting the boundaries of raw beef, the outer contour coordinates of the lean meat region may be (100, 50), (300, 250)..., the outer contour coordinates of the fat meat region may be (50, 100), (80, 100)..., and the outer contour coordinates of the fascia region may be (180, 120), (220, 120)... In some possible embodiments, the boundary contour between lean and fat meat may also be extracted, and the boundary contour between lean and fat meat may be (100, 100)... Then these contour coordinates are integrated to generate data on the component distribution boundaries of raw beef meat.

[0097] Understandably, step S2 in this method uses a component extraction strategy to segment and extract the boundaries of different image features of lean meat, fat, and fascia using appropriate algorithms, ensuring the accuracy of the boundaries of each component.

[0098] Step S23: Obtain the total boundary pixel data based on the distribution boundary of the raw meat components, obtain the proportion data of the raw meat components based on the total boundary pixel data, and simultaneously update the proportion data of the raw meat components in real time based on the update mechanism.

[0099] In this embodiment, step S23 includes:

[0100] Step S231: Process the distribution boundary of the original meat components based on the contour tracking algorithm to obtain the number of pixels of lean meat boundary, the number of pixels of fat boundary, and the number of pixels of fascia boundary.

[0101] Specifically, by traversing the contour point by point, the number of pixels of various boundaries in the distribution boundaries of each component of the raw meat is counted, including the number of pixels of lean meat boundary, fat meat boundary, and fascia boundary. The number of pixels can directly reflect the length of various boundaries, and the length of the boundary is positively correlated with the area of ​​the component region. For example, under the same shape, the larger the region, the longer the boundary, which can provide basic data for subsequent proportion calculations.

[0102] For example, the number of pixels for the lean meat outline of the raw beef can be counted point by point, assuming a total of 1200 pixels. The number of pixels for the fat outline of the raw beef can be counted point by point, assuming a total of 300 pixels. The number of pixels for the fascia outline of the raw beef can be counted point by point, assuming a total of 400 pixels. This gives the number of pixels for the three types of boundaries of the raw beef: 1200 for lean meat, 300 for fat, and 400 for fascia.

[0103] Step S232: Obtain the total boundary pixel data based on the lean meat boundary pixel count data, the fat meat boundary pixel count data, and the fascia boundary pixel count data.

[0104] Specifically, the pixels obtained in step S231 are summed to obtain the total boundary pixel data, which provides a reference standard for subsequent component ratio calculations.

[0105] Furthermore, taking the data obtained in step S231 as an example, if the lean meat is 1200, the fat is 300, and the fascia is 400, then the total number of boundary pixels is 1200 + 300 + 400 = 1900.

[0106] Step S233: Calculate and obtain the original meat component ratio data using the total boundary pixel data. The original meat component ratio data includes lean meat component ratio data, fat meat component ratio data, and fascia component ratio data.

[0107] Specifically, using the total number of boundary pixels obtained in step S232 as the denominator and the number of boundary pixels of each type as the numerator, the proportions of lean meat, fat, and fascia are calculated respectively. The calculation formulas are as follows: lean meat proportion = number of lean meat boundary pixels / total number of boundary pixels; fat proportion = number of fat boundary pixels / total number of boundary pixels; fascia proportion = number of fascia boundary pixels / total number of boundary pixels. Through these calculated proportions, the relative proportions of each component in the raw meat are quantified.

[0108] Furthermore, taking the assumed total boundary pixel data of 1900 obtained in step S232, and the assumed lean meat of 1200, fat of 300, and fascia of 400 as examples, the proportion of lean meat is approximately 1200 / 1900 = 63.16%, the proportion of fat is approximately 300 / 1900 = 15.79%, and the proportion of fascia is approximately 400 / 1900 = 21.05%.

[0109] Step S234: When acquiring the raw meat component ratio data, the raw meat component ratio data deviation is obtained through an update mechanism and based on the input raw meat texture enhancement image data, and the raw meat component ratio data is updated in real time based on the raw meat component ratio data deviation.

[0110] Understandably, when acquiring raw meat component ratio data, a continuous monitoring system can be envisioned, capturing one frame of image per second. When a new raw meat texture enhancement image is input, steps S231-S233 are re-executed to calculate the new component ratio. The new component ratio data is compared with the historical ratio to calculate the ratio deviation, i.e., the new component ratio data minus the historical ratio data. If the absolute value of the deviation exceeds a preset threshold, such as 5%, it can be adjusted according to actual processing needs. In this case, the new ratio replaces the historical ratio to complete the update. If the deviation does not exceed the preset threshold, the historical ratio continues to be used.

[0111] For example, assuming the set ratio threshold is 5%, and a new enhanced image of raw beef texture is input, the recalculated lean meat (1150), fat (320), and fascia (410) pixels result in a total boundary pixel count of 1880. The new ratio is approximately 61.17% lean meat, 17.02% fat, and 21.81% fascia. The deviation between the new ratio and the historical ratio is calculated, using the ratio data from step S233. The slice deviation is -1.99% lean meat, 1.23% fat, and 0.76% fascia. Since the absolute values ​​do not exceed 5%, the historical ratio can continue to be used for subsequent steps. However, if a new enhanced image of raw beef is input, and the recalculated ratio is approximately 28.57% fascia, the deviation from the historical ratio is 7.52%, and the absolute value of the deviation exceeds 5%. In this case, the fascia ratio is updated to 28.57%, and other ratios are updated simultaneously.

[0112] Understandably, step S23 can quantify the number of pixels at the boundaries of the raw meat components, calculate the proportion of each component, and ensure that the proportion data can dynamically reflect the actual state of the raw meat through a real-time update mechanism. If the raw meat moves, the proportion deviation can be corrected in time to ensure the timeliness of the data. The number of boundary pixels reflects the length of the component boundary, which is indirectly related to the area. The proportion calculation can use the sum of boundary pixels as a benchmark to calculate the proportion, ensuring the objectivity of the result.

[0113] Step S24: Based on the proportion data of fascia components in the original meat component proportion data, a proportion judgment is made to determine the low fascia location data.

[0114] In this embodiment, step S24 includes:

[0115] Step S241: If the proportion of fascia components is greater than or equal to a predefined fascia threshold, obtain the pixel density data of the fascia region based on the original meat component distribution boundary, and determine the range of the low fascia region based on the pixel density data of the fascia region.

[0116] Specifically, when the proportion of fascia components is greater than or equal to a predefined fascia threshold, it indicates that the fascia content of the raw meat is high. It is necessary to actively find areas with low fascia density. Based on the distribution boundary of the raw meat components, the coordinate range of all fascia regions can be extracted, and then the pixel density of each fascia region can be calculated. The pixel density can be expressed as the number of fascia pixels in the fascia region / the total area of ​​the region. The lower the pixel density, the less fascia content in the region, and vice versa. The fascia region with the lowest pixel density is defined as the low fascia region range.

[0117] Understandably, a predefined fascia threshold can be set, such as 20%, but in actual processing, it can be set according to the processing requirements of the meat. The higher the threshold, the lower the tolerance for fascia content.

[0118] For example, assuming a predefined threshold of 20%, the fascia ratio of raw beef is 21.05%. Since 21.05% is greater than 20%, it belongs to the high fascia ratio. Based on the component distribution boundary, two fascia regions are extracted, such as region 1 (X-axis 180-220, Y-axis 120-160, area of ​​1600 pixels) and region 2 (X-axis 250-280, Y-axis 80-110, area of ​​900 pixels). Then the pixel density can be calculated. Region 1 has 800 fascia pixels, density = 800 / 1600 = 0.5. Region 2 has 300 fascia pixels, density = 300 / 900 ≈ 0.33. Region 2 has a lower density, so region 2 is defined as the low fascia region range.

[0119] Step S242: Obtain feature point set data of the low fascia region based on the range of the low fascia region, group the feature point set data of the low fascia region, obtain pixel distribution data of the region range, and confirm the center data of the low fascia location based on the pixel distribution data of the region range.

[0120] Specifically, the process involves extracting and grouping feature points, then analyzing pixel distribution to determine the location center data. The Harris corner detection algorithm can be used to extract feature point sets within the low fascia region, such as the corner points at the edge of the region and pixels with significant grayscale changes. These points reflect the geometric features of the region. Subsequently, these feature point sets can be grouped according to coordinates, and the number of pixels in each group can be counted. The group with the most concentrated pixel distribution is selected, which represents the sub-region with the most stable features. Then, the geometric center of the coordinate range of this group is calculated to obtain the location center data of the low fascia.

[0121] Understandably, the Harris focus detection algorithm is a corner detection method based on image grayscale changes. It can identify points with significant grayscale changes in multiple directions, i.e., corners, by analyzing the grayscale changes in local areas of an image. In this embodiment, low fascia areas often have complex texture changes. The Harris focus detection algorithm can capture these texture changes more accurately and extract corners that reflect the geometric features of the area. Furthermore, raw meat processing requires real-time image analysis and the generation of cutting paths. The Harris focus detection algorithm can avoid complex feature value solutions and meet the real-time requirements of raw meat processing.

[0122] For example, the Harris corner detection algorithm is applied to the low fascia region (X-axis 250-280, Y-axis 80-110) of raw beef to extract 20 feature points, such as (255, 85) and (260, 90). The feature points are divided into 3 groups according to the X coordinate: group 1 (X-axis 250-260, containing 6 points), group 2 (X-axis 260-270, containing 8 points), and group 3 (X-axis 270-280, containing 6 points). The pixel distribution of group 2 is found to be the most concentrated, containing 8 points, with X-axis 260-270 and Y-axis 85-95. The geometric center of this group is calculated as X = (260+270) / 2 = 265, y = (85+95) / 2 = 90. The low fascia center data can be preliminarily determined to be (265, 90).

[0123] Step S243: Optimize the low fascia location center data to obtain optimized low fascia location center data; calculate the region pixel mean data and region pixel variance data based on the fascia region pixel density data; and obtain low fascia location data through the optimized low fascia location center data, the region pixel mean data, and the region pixel variance data.

[0124] Specifically, outliers in the feature point set of the group with determined fascia location data are removed, such as isolated points that deviate from the distribution set. The geometric center of the group is then recalculated to obtain optimized low fascia location center data. Next, the mean and variance of the region pixels are calculated. The mean is the average gray value of all pixels in the region, and the variance is the dispersion of the gray values. The smaller the variance, the more uniform the gray value in the region and the more stable the fascia distribution. Finally, the optimized center coordinates, mean, and variance are integrated to obtain low fascia location data. The mean and variance data can be used to verify the stability of the region.

[0125] For example, for 20 feature points in the low fascia region of raw beef, after removing two outlier points such as (250, 80) and (280, 110), 8 points are still included in group 2 of the remaining 18 points. The X-axis and Y-axis ranges of group 2 are recalculated, and the geometric center can be optimized from (265, 90) to (264, 89), which is closer to the actual pixel distribution. Then, the mean and variance of pixel grayscale in this region are calculated. For example, the mean is 75 and the variance is 12. Finally, the low fascia location data is obtained as the center coordinates (264, 89), the mean pixel value of the region is 75, and the variance of the region is 12.

[0126] Step S244: If the proportion of fascia components is less than a predefined fascia threshold, then low fascia location data is obtained based on the original meat component distribution boundary.

[0127] Understandably, when the proportion of fascia components is less than the predefined fascia threshold, it indicates that the overall fascia content of the raw meat is low. Without the above analysis, the low fascia location can be determined directly based on the distribution boundary of the raw meat components, that is, the geometric center can be extracted from the distribution boundary of the raw meat components.

[0128] Step S3: Based on the low fascia location data, an initial raw meat cutting path is obtained, and a path adjustment strategy is performed on the initial cutting path to obtain the final raw meat cutting path.

[0129] In this embodiment, step S3 includes:

[0130] Step S31: Generate the initial raw meat cutting path based on the A* search algorithm.

[0131] The A-star search algorithm takes low fascia location data as input and uses this data as the starting point to perform path search on the original meat texture enhancement image data.

[0132] Understandably, the A* search algorithm searches for the initial raw meat cutting path based on a cost function and a heuristic function. It uses the center coordinates in the low fascia location data as the starting point of the path and sets the target point as the edge of the raw meat. For example, if the processing requirement is longitudinal cutting, the target point is the center of the right or left edge of the raw meat. The A* search algorithm searches in the pixel grid of the raw meat texture enhancement image to generate the initial raw meat cutting path from the starting point to the target point.

[0133] Understandably, the cost function used by the A* search algorithm can be simply understood as the actual length of the current path, such as the number of pixels already traversed, while the heuristic function used by the A* search algorithm can be simply understood as the estimated distance from the current point to the target point, such as Euclidean distance, Manhattan distance, etc.

[0134] For example, suppose the raw beef needs to be cut longitudinally from left to right. Let's set the input parameters for the A* algorithm: the starting point is the center of the low fascia location (264, 89), and the target point is the center of the right edge of the raw beef (350, 150). Based on the component distribution boundary, the right edge X=350 is determined.

[0135] Furthermore, path search is performed using the cost function F(n) = g(n) + h(n), where g(n) represents the number of pixels along the path from the starting point to the current point n, and h(n) is the Euclidean distance from the current point n to the target point. The A* algorithm searches within the pixel grid of the beef raw meat texture enhancement image until the initial raw meat cutting path is generated, such as (264,89)→(280,100)→(300,120)→(320,135)→(350,150).

[0136] Step S32: Obtain raw meat component distribution data, and perform a path adjustment strategy on the initial raw meat cutting path based on the raw meat component distribution data to obtain the final raw meat cutting path.

[0137] In this embodiment, step S32 includes:

[0138] Step S321: Obtain raw meat component distribution data based on the raw meat texture enhancement image and the raw meat component distribution boundary, and simultaneously obtain high fascia location data based on the low fascia location data.

[0139] Understandably, by summarizing the coordinate range and boundary data of the fat, lean meat, and fascia regions, the spatial location of each component is clarified. At the same time, the high fascia location is determined. The fascia region where the proportion of fascia components is greater than the preset high fascia threshold is located as the high fascia location. This location is an obstacle that needs to be avoided when adjusting the initial raw meat cutting path.

[0140] For example, assuming the set high fascia threshold is 30%, areas exceeding the high fascia threshold are considered to have a high fascia content. If the fascia ratio of region 1 is 50% and that of region 2 is 33.3%, and both regions exceed the set high fascia threshold, then both fascia regions are defined as high fascia locations.

[0141] Step S322: Determine the fat separation area based on the original meat component distribution data, and obtain obstacle coordinate data based on the high fascia location data and the obstacle acquisition mechanism.

[0142] Understandably, the fat region is defined as the fat separation region. Since it needs to be cut and separated from the lean meat, the path needs to avoid the fat region to avoid cutting into the fat. At the same time, the high fascia location is regarded as a path obstacle. The obstacle is quantified by the obstacle center + obstacle range method. The obstacle center is the geometric center of the high fascia region, and the obstacle range is the coordinate range of the high fascia region. The two together constitute the obstacle coordinate data.

[0143] For example, if the fat region is (X-axis 50-80, Y-axis 100-200) and (X-axis 320-350, Y-axis 80-180), then the fat region is defined as the fat separation region. If the high fascia region is (X-axis 180-220, Y-axis 120-160) and (X-axis 250-280, Y-axis 80-110), then the obstacle coordinate data is calculated. The center of obstacle 1 is (200, 140), and the range is X-axis 180-220, Y-axis 120-160. The center of obstacle 2 is (265, 95), and the range is X-axis 250-280, Y-axis 80-110. The integrated obstacle coordinate data is [(200, 140), (180-220, 120-160)] and [(265, 95), (250-280, 80-110)].

[0144] Step S323: The initial raw meat cutting path is corrected and adjusted using the obstacle coordinate data and the fat separation area to determine the final raw meat cutting path, i.e., the final raw meat cutting path.

[0145] Understandably, the positional relationship between obstacle coordinate data, fat separation area and initial raw meat cutting path is analyzed. If a point on the initial path enters the obstacle range or fat separation area, the distance between that point and the center of the obstacle and the center of the fat area is calculated. Then, the deviation points on the path can be adjusted by methods such as artificial potential field method to make them move away from the obstacle and fat area, thereby generating the final raw meat cutting path.

[0146] For example, using the initial cutting path of the raw beef (264,89)→(280,100)→(300,120)→(320,135)→(350,150), where (320,135) is located at the edge of the fat separation area (X-axis 320-350, Y-axis 80-180), there is a risk of cutting into the fat. Therefore, the artificial potential field method can be used to treat the obstacle and the fat area as a repulsive force field, pushing the path away. The repulsive force pushes the point to the left by 10 pixels, adjusting it to (310,140). The adjusted path is verified to be (264,89)→(280,100)→(300,120)→(310,140)→(350,150). This path completely avoids the high fascia obstacle and the fat separation area, meets the cutting conditions, and can be finally determined as the final cutting path of the raw meat.

[0147] Step S4: Obtain the final cutting instruction based on the final cutting path of the raw meat, and drive the cutting module to perform the cutting operation according to the final cutting instruction.

[0148] Understandably, the final cutting path of the raw meat is converted into machine instructions that the cutting module can recognize, and the drive module drives the cutting module to perform the final precise cutting of the raw meat based on the machine instructions.

[0149] Figure 3 The diagram shows a schematic of a meat product processing control system that can realize the ideas of this application, according to some embodiments of this application. The following is a detailed description of this meat product processing control system.

[0150] Specifically, a meat product processing control system includes:

[0151] The enhancement module is used to acquire raw meat surface image data and raw meat stability feature image data;

[0152] A boundary confirmation module is used to acquire enhanced image data of the original meat texture and determine the distribution boundary of the original meat components;

[0153] The location confirmation module is used to perform proportion extraction to obtain raw meat component proportion data and low fascia location data.

[0154] The instruction generation module is used to obtain an initial cutting path, execute a path adjustment strategy to obtain the final cutting path of the raw meat, and obtain the final cutting instruction based on the final cutting path of the raw meat.

[0155] A cutting module, which is used to cut the raw meat;

[0156] A driving module is used to drive the cutting module to perform a cutting operation according to the final cutting command.

[0157] The specific usage and function of this embodiment are explained below:

[0158] Based on the above, this application embodiment first acquires raw meat surface image data, performs enhanced preprocessing on the raw meat surface image data to acquire stable feature image data of raw meat, and eliminates interference information in the raw meat surface image, such as reflection, uneven brightness, etc. Then, based on the stable feature image data of raw meat, it acquires raw meat texture enhanced image data and determines the raw meat component distribution boundary. The raw meat component distribution boundary is proportionally extracted to acquire raw meat component proportion data. Based on the raw meat component proportion data, it acquires low fascia position data to provide coordinate basis for subsequent cutting path planning. Then, based on the low fascia position data, it performs an initial raw meat cutting path and acquires the initial cutting path. The initial cutting path is then adjusted using a path adjustment strategy to acquire the final raw meat cutting path. Finally, based on the final raw meat cutting path, it acquires the final cutting command and drives the cutting module to perform the cutting operation according to the final cutting command. Through multi-level image analysis and dynamic path planning, it achieves accurate separation of raw meat and maximizes the utilization rate of raw meat. It provides a meat product processing control method that can enhance color recognition and ensure accurate separation of raw meat by the cutting process based on color recognition.

[0159] This application also provides a schematic diagram of an electronic device that can implement the concept of this application, and the electronic device will be described in detail below.

[0160] Specifically, an electronic device includes:

[0161] At least one processor; and at least one memory communicatively connected to the processor; wherein the memory stores instructions executable by at least one processor, the instructions being executed by at least one processor to enable at least one processor to perform the method proposed in Embodiment 1 of the present invention.

[0162] The following is a detailed introduction to the various components of the electronic device:

[0163] In this context, the processor is the control center of the electronic device. It can be a single processor or a collective term for multiple processing elements. For example, a processor can be one or more central processing units (CPUs), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement Embodiment 1 of this invention, such as one or more digital signal processors (DSPs) or one or more field-programmable gate arrays (FPGAs).

[0164] The processor can perform various functions of an electronic device by running or executing software programs stored in memory and by calling data stored in memory.

[0165] The memory is used to store the software program that executes the solution of the present invention, and the execution is controlled by the processor. For specific implementation methods, please refer to the above method embodiments, which will not be repeated here.

[0166] The memory can be a real-only memory (ROM) or other type of static storage device capable of storing static information and instructions, a random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only (CD-ROM), or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media, or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto. The memory can be integrated with the processor or exist independently and coupled to the processor through an interface circuit of an electronic device; this embodiment of the invention does not specifically limit this.

[0167] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via limited means (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0168] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0169] It should be understood that, in the embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0170] The above-described 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 the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for controlling the processing of meat products, characterized in that, The method includes: Obtain raw meat surface image data, perform enhancement preprocessing on the raw meat surface image data, and obtain raw meat stable feature image data; Based on the original meat stable feature image data, obtain the original meat texture enhancement image data, determine the original meat component distribution boundary, extract the proportion of the original meat component distribution boundary to obtain the original meat component proportion data, and obtain the low fascia location data based on the original meat component proportion data, including: Brightness is adjusted based on gamma correction of the original meat stable feature image to obtain contrast-enhanced image data of the original meat. Simultaneously, texture enhancement is performed on the contrast-enhanced image of the original meat to obtain texture-enhanced image data of the original meat. Boundary extraction is performed on the texture-enhanced image data of the original meat using a component extraction strategy to obtain the distribution boundaries of the original meat components. Based on the distribution boundaries of the original meat components, the total boundary pixel data is obtained, and the proportion data of the original meat components is obtained based on the total boundary pixel data. Simultaneously, the proportion data of the original meat components is updated in real time based on an update mechanism. Based on the proportion data of the fascia components in the proportion data of the original meat components, a proportion judgment is performed to determine the low fascia location data. Specifically, the lean meat component region in the enhanced meat texture image data is segmented using a clustering algorithm to obtain lean meat component region data; the fat component region in the enhanced meat texture image data is segmented using a threshold segmentation algorithm to obtain fat component region data; the fascia component region in the enhanced meat texture image data is segmented using a Canny edge detection algorithm to obtain fascia component region data; contour detection is performed on the lean meat component region data, the fat component region data, and the fascia component region data to obtain lean-fat-tendon boundary data; and the distribution boundary of the original meat components is obtained based on the lean-fat-tendon boundary data. An initial raw meat cutting path is obtained based on low fascia location data. Then, a path adjustment strategy is applied to this initial cutting path to obtain the final raw meat cutting path, including: An initial raw meat cutting path is generated based on the A-Star search algorithm; wherein, the input of the A-Star search algorithm is low fascia location data, and the low fascia location data is used as the starting point for the A-Star search algorithm to perform path search on the raw meat texture enhancement image data; raw meat component distribution data is obtained, and a path adjustment strategy is executed on the initial raw meat cutting path based on the raw meat component distribution data to obtain the final raw meat cutting path; The final cutting instruction is obtained based on the final cutting path of the raw meat, and the cutting module is driven to perform the cutting operation according to the final cutting instruction.

2. A meat product processing process control method according to claim 1, characterized in that, Acquire raw meat surface image data, perform enhancement preprocessing on the raw meat surface image data, and obtain raw meat stable feature image data, including: The raw meat is photographed in real time using a camera to obtain surface image data of the raw meat. Based on the surface image data of the raw meat, pixel histogram data of the raw meat is obtained, and pixel distribution data of the raw meat is obtained based on the pixel histogram data of the raw meat. The original meat pixel distribution data is balanced and adjusted using a gray-level cumulative distribution function to obtain balanced image data of the original meat. The gray-level range is determined based on the balanced image data of the original meat, and the extended image data of the original meat is obtained through the gray-level range. Color analysis is performed on the extended image data of the raw meat to extract the color distribution data of the raw meat. Stability analysis is then performed on the color distribution data of the raw meat to obtain stable feature image data of the raw meat.

3. A meat product processing process control method according to claim 1, characterized in that, Based on the distribution boundary of the raw meat components, the total boundary pixel data is obtained; based on the total boundary pixel data, the proportion data of the raw meat components is obtained; and simultaneously, the proportion data of the raw meat components is updated in real time based on an update mechanism, including: The boundary of the original meat component distribution is processed based on the contour tracking algorithm to obtain the number of pixels of lean meat boundary, the number of pixels of fat boundary, and the number of pixels of fascia boundary. The total number of boundary pixels is obtained based on the number of pixels at the boundaries of lean meat, fat, and fascia. The original meat component ratio data is obtained by calculating the total boundary pixel data. The original meat component ratio data includes lean meat component ratio data, fat meat component ratio data and fascia component ratio data. When acquiring the raw meat component ratio data, an update mechanism is used to obtain the raw meat component ratio data deviation based on the input raw meat texture enhancement image data, and the raw meat component ratio data is updated in real time based on the raw meat component ratio data deviation.

4. A meat product processing process control method according to claim 1, characterized in that, Based on the proportion of fascia components in the original meat composition data, a proportion judgment is made to determine the low fascia location data, including: If the proportion of the fascia components is greater than or equal to a predefined fascia threshold, then the pixel density data of the fascia region is obtained based on the original meat component distribution boundary, and the range of the low fascia region is determined based on the pixel density data of the fascia region. Based on the range of the low fascia region, obtain the feature point set data of the low fascia region, group the feature point set data of the low fascia region, obtain the pixel distribution data of the region range, and confirm the center data of the low fascia location based on the pixel distribution data of the region range. The low fascia location center data is optimized to obtain optimized low fascia location center data. The mean pixel data and variance data of the region are calculated based on the pixel density data of the fascia region. The low fascia location data is obtained through the optimized low fascia location center data, the mean pixel data of the region, and the variance data of the region. If the proportion of fascia components is less than a predefined fascia threshold, then low fascia location data are obtained based on the original meat component distribution boundary.

5. A meat product processing process control method according to claim 1, characterized in that, Obtain raw meat component distribution data, and based on the raw meat component distribution data, execute a path adjustment strategy on the initial raw meat cutting path to obtain the final raw meat cutting path, including: Based on the enhanced image of the original meat texture and the boundary of the original meat component distribution, the original meat component distribution data is obtained, and at the same time, the high fascia location data is obtained based on the low fascia location data. The fat separation area is determined based on the original meat component distribution data, and obstacle coordinate data is obtained based on the high fascia location data and the obstacle acquisition mechanism. The initial raw meat cutting path is corrected and adjusted by using the obstacle coordinate data and the fat separation area to determine the final raw meat cutting path, i.e., the final raw meat cutting path.

6. A meat processing process control system applying the method according to any one of claims 1 to 5, characterized in that The system includes: The enhancement module is used to acquire raw meat surface image data and raw meat stability feature image data; A boundary confirmation module is used to acquire enhanced image data of the original meat texture and determine the distribution boundary of the original meat components; The location confirmation module is used to perform proportion extraction to obtain raw meat component proportion data and low fascia location data. The instruction generation module is used to obtain an initial cutting path, execute a path adjustment strategy to obtain the final cutting path of the raw meat, and obtain the final cutting instruction based on the final cutting path of the raw meat. A cutting module, which is used to cut the raw meat; A driving module is used to drive the cutting module to perform a cutting operation according to the final cutting command.

Images