Meat grading method and system

Abstract

The present application provides a meat grading method and system. The method comprises: acquiring multi-dimensional meat imaging information of meat; performing high-precision measurement on the meat to obtain grading truth value data of the meat; and computing a grading index parameter computation value on the basis of the multi-dimensional meat imaging information, and training a grading index parameter model on the basis of the grading index parameter computation value and the grading truth value data to obtain a trained grading index parameter model.

Description

Meat grading method and system

[0001] This application claims priority to Chinese patent application No. 202311195747.2 filed on September 15, 2023, the entire contents of which are incorporated by reference into this application. Technical Field

[0002] The present disclosure relates to the field of detection technology, and more particularly to a meat grading method and system, and more particularly to a meat grading method and system based on model training. Background Art

[0003] During the production and processing of meats like pork, lamb, and beef, the meat undergoes grading and processing. The former uses various carcass parameters to determine the grade of the meat, thereby settling accounts with upstream factories. The latter involves cutting the meat carcass into different parts or regions, which are then sold at different values. For meat manufacturers, the higher the accuracy of the grading process, the greater the economic benefits. Therefore, pursuing new technologies to improve grading is a consistent and long-term pursuit for manufacturers. For example, during the grading process, backfat thickness is measured and combined with visual observation to determine the grade. This not only requires a high level of technical expertise from workers and results in high labor costs for manufacturers, but the grading results can also be affected by individual worker experience and performance, resulting in uncertainty about the stability and consistency of grading results over time.

[0004] Overall, the meat grading process relies heavily on manual labor, which not only incurs high labor costs but also makes it difficult to ensure accuracy, stability, and consistency during the grading process.

[0005] Summary of the Invention

[0006] According to one aspect of the present disclosure, a meat grading method is provided, the method comprising: acquiring multidimensional meat imaging information of a meat piece; performing high-precision measurement on the meat piece to acquire grading true value data of the meat piece; calculating grading index parameter calculated values ​​based on the multidimensional meat imaging information, and training a grading index parameter model based on the grading index parameter calculated values ​​and the grading true value data to obtain a trained grading index parameter model.

[0007] In an embodiment, the meat pieces are graded according to the grading index parameter calculation values ​​calculated by the trained grading index parameter model.

[0008] In an embodiment, the calculated value of the grading index parameter includes at least the morphological structure information and composition information of the meat piece as a whole and at least one part, and the method also includes grading the meat piece according to the morphological structure information and composition information required for grading and referring to the grading standard.

[0009] In an embodiment, the calculated value of the grading index parameter includes division information of at least one part of the meat piece and position information of at least one part.

[0010] In an embodiment, the morphological structure information includes at least one of shape information, length information, width information, thickness information, and weight information of the meat piece as a whole and each of at least one part; the composition information includes at least one of mass percentage, lean meat rate, and fat rate of the meat piece as a whole and each of at least one part.

[0011] In an embodiment, training the grading indicator parameter model includes: calculating the difference between the grading indicator parameter calculated value and the grading true value data, and training the grading indicator parameter model by comparing the difference with corresponding reference data.

[0012] In an embodiment, training the grading indicator parameter model also includes: when the difference is greater than the reference data, updating the model parameters of the grading indicator parameter model to recalculate the grading indicator parameter calculated value, and comparing the difference between the recalculated grading indicator parameter calculated value and the grading true value data with the reference data, repeating the calculation and comparison until the difference is less than or equal to the reference data.

[0013] In an embodiment, using the grading index parameter model to calculate the grading index parameter calculated value includes: the grading index parameter model performs image processing on the meat imaging information to divide the meat imaging information, and calculates the grading index parameter calculated value based on the divided meat imaging information.

[0014] In an embodiment, using the grading index parameter model to calculate the grading index parameter calculated value includes: the grading index parameter model receives meat imaging information that has been divided through image processing, and calculates the grading index parameter calculated value based on the divided meat imaging information.

[0015] In an embodiment, the image processing includes at least one of image segmentation and positioning processing, image noise reduction processing, and image enhancement processing.

[0016] In an embodiment, an image segmentation and positioning model is trained using the meat imaging information and a portion of the grading truth data, and a portion of the grading index parameter calculation value is provided based on the meat imaging information using the output of the trained image segmentation and positioning model.

[0017] In an embodiment, the high-precision measurement includes obtaining the true grading data of the meat piece using a high-precision imaging device; and / or obtaining the true grading data of the meat piece by dividing the meat piece to obtain at least one part and measuring the at least one part separately.

[0018] According to another aspect of the present disclosure, a meat grading system is provided, which executes the meat grading method according to aspects of the present disclosure, and the system includes: a meat imaging detection device, which obtains multi-dimensional meat imaging information of a meat piece; a true value data acquisition device, which obtains grading true value data of the meat piece; a grading index parameter model, which calculates grading index parameter calculated values ​​based on the multi-dimensional meat imaging information, and is trained based on the grading index parameter calculated values ​​and the grading true value data to obtain a trained grading index parameter model.

[0019] In an embodiment, the trained grading index parameter model grades the meat pieces according to the grading index parameter calculation values ​​calculated by the trained grading index parameter model.

[0020] In an embodiment, the system also includes an image segmentation and positioning model, which is trained based on the meat imaging information and a portion of the grading true value data, and uses the trained image segmentation and positioning model to provide an output based on the meat imaging information as part of the grading index parameter calculation value.

[0021] In an embodiment, the graded true value data acquisition device includes: a manual measuring device, which is used to divide the meat piece and directly measure at least one divided part to obtain graded true value data; and / or a high-precision imaging device, which obtains high-precision imaging information of the meat piece with a resolution higher than that of the meat imaging detection device, and obtains the graded true value data therefrom.

[0022] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure, wherein:

[0024] FIG1 illustrates a meat grading system according to an embodiment of the present disclosure.

[0025] FIG2 shows an example of calculated values ​​of grading index parameters according to an embodiment of the present disclosure.

[0026] FIG3 shows a training flowchart of a hierarchical index parameter model according to an embodiment of the present disclosure.

[0027] FIG4 shows a training flowchart of a hierarchical indicator parameter model according to another embodiment of the present disclosure.

[0028] FIG5 shows a schematic diagram of part division of meat imaging information according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0029] To more clearly illustrate the objectives, technical solutions, and advantages of the present disclosure, embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings. It should be understood that the following description of the embodiments is intended to explain and illustrate the overall concept of the present disclosure and should not be construed as limiting the present disclosure. In the specification and drawings, the same or similar reference numerals refer to the same or similar parts or components. For the sake of clarity, the drawings are not necessarily drawn to scale, and some well-known parts and structures may be omitted in the drawings.

[0030] Unless otherwise defined, technical or scientific terms used in this disclosure should have the ordinary meaning understood by a person of ordinary skill in the art to which this disclosure belongs. The terms "first," "second," and similar terms used in this disclosure do not denote any order, quantity, or importance, but are simply used to distinguish different components. The terms "a" or "an" do not exclude a plurality. Terms such as "include" or "comprise" mean that the element or object preceding the term includes the elements or objects listed after the term and their equivalents, but do not exclude other elements or objects. Terms such as "connected" or "connected" are not limited to physical or mechanical connections but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," "right," "top," or "bottom" are used only to indicate relative positional relationships; if the absolute position of the described objects changes, the relative positional relationship may also change accordingly. When an element such as a layer, film, region, or substrate is referred to as being "on" or "under" another element, the element may be "directly" "on" or "under" the other element, or intervening elements may be present.

[0031] The purpose of the present disclosure is to propose a meat grading method and system, which effectively improves the accuracy of the meat grading process by training corresponding models according to meat grading requirements, thereby reducing the losses of meat manufacturers and improving economic benefits.

[0032] FIG1 shows a meat grading system 1000 according to an embodiment of the present disclosure. As shown in FIG1 , the meat grading system 1000 may include a meat imaging detection device 10 , a true value data acquisition device 20 , and a grading index parameter model 30 .

[0033] As shown in Figure 1, the meat imaging detection device 10 can be configured to detect meat pieces to obtain multi-dimensional meat imaging information MD of the meat pieces. In this article, "meat imaging information" can refer to a multi-dimensional (for example, three-dimensional) image of meat and the tissue composition and part area information included in the image. According to an embodiment of the present disclosure, by using the meat imaging detection device 10 as an automated imaging detection means, compared with traditional manual grading and processing, the surface and internal information of meat can be more comprehensively and deeply mined. In addition, in the obtained meat imaging information MD, different tissue components in the meat, such as bones, fat and lean meat, as well as different parts and areas such as the front tip segment, middle segment (including ribs), and back tip segment can be clearly seen, and most of these image information cannot be observed by the human eye.

[0034] As an example, the meat imaging detection device 10 may include any one of an X-ray imaging device, a visible light imaging device, an infrared imaging device, and an ultrasonic imaging device, or any combination thereof. As an example, the meat imaging information MD may include any one of an electronic computed tomography (CT) image, a digital imaging technology (DR) image, and the like. However, the embodiments of the present disclosure are not limited thereto.

[0035] In this way, by using meat imaging detection methods to obtain meat imaging information MD as the basis for calculating processing parameters, in addition to ensuring the accuracy of calculation and processing, the use of meat imaging detection methods can also obtain richer and more multi-dimensional image information compared to human eye observation. For example, X-ray CT imaging can obtain material and structural information inside the meat, which is not directly observable by the human eye, thereby providing more information for the meat grading and processing process, which is conducive to promoting its accuracy.

[0036] As shown in Figure 1, the true value data acquisition device 20 can be configured to perform high-precision measurements on meat pieces to obtain grading true value data TD of the meat pieces. In this article, "grading true value data" can refer to data that enables a device with a grading function to achieve grading of meat pieces with the required accuracy. That is, when using the graded true value data, the desired grading result can be obtained. Therefore, according to an embodiment of the present disclosure, the "grading true value data" can be used as reference data for training the grading index parameter model 30, and can be data of the same type as the grading index parameter calculated value calculated by the grading index parameter model 30. In other words, the grading index parameter model 30 can be trained with reference to the grading true value data TD, so that the grading index parameter calculated value calculated by the trained grading index parameter model 30 has an accuracy and reliability comparable to the grading true value data TD.

[0037] As an example, obtaining the graded truth data TD may include dividing the meat piece to obtain at least one part, and separately measuring the at least one part to obtain the graded truth data TD of the meat piece. In this case, the truth data acquisition device 20 may include a manual measuring device. The manual measuring device can be used to divide the meat piece and directly measure at least one divided part to obtain the graded truth data TD. For example, an experienced meat processor may first cut the meat piece into one or more parts using a specific manual measuring device, and then directly measure each part one by one. For example, in this way, the morphological structure information such as the length information and width information of the entire meat piece can be measured. For another example, the proportion information of the cut part, such as the lean meat rate, etc., can be measured. In this way, through direct manual measurement, accurate and reliable measurement results can be obtained as graded truth data TD, which can be used for subsequent training processes.

[0038] As another example, the true value data acquisition device 20 may include a high-precision imaging device. The high-precision imaging device may be a device with a higher resolution than the meat imaging detection device 10. For example, the high-precision imaging device may have higher performance indicators, such as spatial resolution, material resolution, etc., compared to the meat imaging detection device 10. In this way, after obtaining high-precision imaging information, the difficulty of image segmentation and positioning is lower, and thus the accuracy is higher, so that more accurate segmentation and positioning data can be obtained as true value data. As an example, the high-precision imaging device may include a high-performance medical CT device. In addition, as an example, the processing of high-precision imaging information may include automatic computational processing (for example, using a threshold segmentation algorithm), or manual or semi-automatic image processing, and the like.

[0039] As another example, the true value data acquisition device 20 may also include an image processing device. In some cases, the resolution of the meat imaging information MD obtained using the meat imaging detection device 10 may be sufficient. In this case, an image processing device can be used to perform image processing on the meat imaging information MD, for example, automatic processing, semi-automatic processing or manual processing, to obtain part division information and position information. Then, based on this information, the morphological structure information of the meat piece, such as length information, width information, etc., can be calculated manually by a technician or automatically by a computing device. In this way, accurate data can also be obtained as graded true value data TD for subsequent training processes. As an example, the image processing device can be a trained image division and positioning model.

[0040] It should be noted that the above-mentioned true value data acquisition device 20 is merely an example, and the true value data acquisition device of the present disclosure may include any relevant device as long as the device can acquire reference data for training the hierarchical index parameter model 30.

[0041] According to an embodiment of the present disclosure, as shown in FIG1 , a grading index parameter model 30 can receive meat imaging information MD and ground truth data TD from a meat imaging detection device 10 and a ground truth data acquisition device 20, respectively. The grading index parameter model 30 can calculate a grading index parameter calculated value DCR for grading meat pieces based on the multi-dimensional meat imaging information MD, and train the grading index parameter model 30 based on the calculated grading index parameter value DCR and the ground truth data TD to obtain a trained grading index parameter model 30.

[0042] Thereafter, the trained grading index parameter model 30 can be used to automatically grade batches of meat pieces. Specifically, the trained grading index parameter model 30 can be used to calculate a grading index parameter calculated value DCR based on the meat imaging information MD of each meat piece to be graded, and the meat pieces can be graded based on the calculated grading index parameter calculated value DCR for each meat piece.

[0043] As shown in FIG1 , as an example, the trained grading index parameter model 30 can be configured to have a grading function. For example, the trained grading index parameter model 30 can be internally provided with a grading unit. A grading standard reference table can be stored in the grading unit. After the grading index parameter calculated value DCR' of the meat piece is calculated using the trained grading index parameter model 30, the grading unit in the trained grading index parameter model 30 can compare the calculated grading index parameter calculated value DCR' with the stored grading standard reference table to determine the grading result and can output the grading result to the relevant inspection personnel to complete the grading operation of the single meat piece.

[0044] Alternatively, as another example, the trained grading index parameter model 30 may not have a grading function, but the grading operation may be completed by other devices outside the trained grading index parameter model 30. In this case, after the trained grading index parameter model 30 calculates and obtains the grading index parameter calculation value DCR', it can be directly output to other external devices, and the meat grading process is completed by other devices. Similar to the above, other devices may also store a grading standard reference table, and the grading results can be output after comparison. As an example, the other device may be a remote computer storing a grading function program, etc. However, the embodiments of the present disclosure are not limited to this. In addition, the embodiments of the present disclosure are not limited to the above-mentioned grading implementation method, but any suitable grading method based on indicator parameters may be used.

[0045] According to an embodiment of the present disclosure, the trained grading index parameter model 30 calculates the grading index parameter calculated value DCR' based on the multidimensional meat imaging information MD. Since the multidimensional meat imaging information MD is in the form of an image, the meat imaging information MD may, for example, include the following information: coordinate information of specific points on the contour edge or inside the multidimensional image of the meat; tissue distribution within the image (for example, the distribution of lean meat, fat, bones, etc.); regional distribution of meat pieces (for example, the front tip section, middle section, back tip section, head, back, etc. of the meat piece). In this case, the trained grading index parameter model 30 may be configured to obtain the above information from the meat imaging information MD, and calculate the information to obtain the grading index parameter calculated value DCR'.

[0046] The grading index parameter calculated value DCR′ may include information related to the entire meat piece. For example, the grading index parameter calculated value DCR′ may include morphological structure information and composition information of the entire meat piece.

[0047] The grading index parameter calculation value DCR' may also include information related to the parts of the meat piece. Generally, the grading of meat pieces requires reference not only to the overall index of the meat piece, but also to the indexes of multiple parts of the meat piece. As described above, the meat imaging information MD can show different parts and regions of the meat piece, so after performing specific image processing means on the meat imaging information MD, the area and part division can be implemented in the meat imaging information MD, thereby obtaining the division information and position information about the parts. Therefore, as an example, the grading index parameter calculation value DCR' may include the division information of at least one part of the meat piece and the position information of at least one part. For example, the part may be at least one of the front tip segment, middle segment, back tip segment, front chest, back hip, rib area, tenderloin area, fat area, lean meat area, bone area, etc. of the meat piece. As shown in Figure 5, Figure 5 shows the division information of the front tip segment, middle segment, and back tip segment of the DR image of the meat piece. However, the embodiments of the present disclosure are not limited to this.

[0048] In this context, "division information" may indicate the division of meat into parts, i.e., the parts of the meat to be graded, so that the calculated grading index parameter value DCR' can include parameter information for the divided parts. In this context, "position information" may indicate the coordinate information of the divided parts. Furthermore, the calculated grading index parameter value DCR' may also include morphological structure information and composition information for each of at least one part of the meat.

[0049] As an example, the morphological structure information may include at least one of the shape information, length information, width information, thickness information, and weight information of the meat piece as a whole and of each part. And as an example, the composition information may include at least one of the mass percentage, lean meat rate, and fat rate of the meat piece as a whole and of each part. As described above, the multi-dimensional meat imaging information MD may include the distribution of different tissues, so that the trained grading index parameter model 30 can calculate the proportion of each tissue in the meat piece, thereby obtaining parameter indicators such as mass percentage, lean meat rate, and fat rate. In addition, the multi-dimensional meat imaging information MD may include the coordinate information of specific points, so that the trained grading index parameter model 30 can calculate parameter indicators such as the length, height, and thickness of the meat piece. Further, in the meat imaging information MD after the parts are divided, the trained grading index parameter model 30 can calculate the morphological structure information (e.g., length, height, thickness, etc.) and composition information (e.g., mass percentage, lean meat rate, fat rate, etc.) of each divided part in a similar manner as described above.

[0050] Figure 2 shows an example of a calculated value DCR' for a grading index parameter according to an embodiment of the present disclosure. For ease of illustration, only the front breast, rear rump, and ribs are shown as example parts, and only length information is shown as an example of morphological structure information, and lean meat percentage is shown as an example of composition information. However, this is merely an example, and the present disclosure may include other grading index parameters.

[0051] As shown in Figure 2, the grading index parameter calculated value DCR' can include the whole body and three parts (i.e., the front chest, the rear hip, and the rib area), as well as the coordinate information, length information, and lean meat rate information corresponding to the whole body and each part. In this case, based on the calculated grading index parameter calculated value DCR', with reference to the grading standard, the grade of the corresponding meat piece can be determined, thereby completing the grading operation. For example, the grading standard can be the grading standard reference table as described above. In the reference table, for example, the various grades and the numerical range of the index parameters corresponding to each grade can be listed. Thus, the grade can be determined based on the parameter values ​​included in the grading index parameter calculated value DCR'.

[0052] According to an embodiment of the present disclosure, the training of the grading index parameter model 30 may include calculating the difference between the grading index parameter calculated value DCR and the grading true value data TD, and training the grading index parameter model 30 by comparing the difference with the corresponding reference data FD. Further, when the difference is greater than the reference data FD, the model parameters of the grading index parameter model 30 are updated to recalculate the grading index parameter calculated value DCR, and compare the difference between the recalculated grading index parameter calculated value DCR and the grading true value data TD with the reference data FD, and repeat the calculation and comparison until the difference is less than or equal to the reference data FD.

[0053] Figure 3 shows a training flowchart of the grading index parameter model 30 according to an embodiment of the present disclosure. As shown in Figure 3, the grading index parameter model 30 can calculate the grading index parameter calculated value DCR for the first time based on the meat imaging information MD. At this time, the grading index parameter model 30 is an untrained model, so the grading index parameter calculated value DCR calculated for the first time cannot meet the expected effect (such as accuracy). In this case, it is necessary to iteratively train the grading index parameter model 30 based on the true value data, and by modifying the model parameters of the grading index parameter model 30 during each iteration, the calculated results can be gradually approached to the true value data. Ultimately, the model parameters of the grading index parameter model 30 can be corrected to be able to calculate the desired grading index parameter calculated value DCR' based on the meat imaging information MD, so as to be used in the subsequent grading process.

[0054] Therefore, as shown in FIG3 , the grading index parameter model 30 can calculate the difference between the first calculated grading index parameter calculated value DCR and the grading true value data TD (expressed as |DCR-TD| in FIG3 ), and then compare the difference with the reference data FD to determine whether the difference is less than or equal to the reference data FD (expressed as |DCR-TD|≤FD? in FIG3 ). If the difference is not less than or equal to (i.e., greater than) the reference data FD (“No” in FIG3 ), the grading index parameter model 30 can update the model parameters and use the updated grading index parameter model 30 to calculate the grading index parameter calculated value DCR again (i.e., a second time). Then, the difference between the second calculated grading index parameter calculated value DCR and the grading true value data TD is calculated, and the above comparison process is repeated, i.e., the second difference is compared with the reference data FD to determine whether the difference is less than or equal to the reference data FD. If the difference is still not less than or equal to (i.e., greater than) the reference data FD (“No” in FIG2 ), the above parameter update and calculation process is repeated until the difference is less than or equal to the reference data FD. When the second difference is less than or equal to the reference data FD ("Yes" in FIG2 ), it can be determined that the training of the grading index parameter model 30 is completed, and the grading index parameter model 30 at this time is determined as the trained grading index parameter model 30.

[0055] In this way, a model capable of calculating multiple grading index parameters for meat can be obtained, and after training the model, an output result with an accuracy close to the ground truth data can be obtained. Therefore, the meat grading system 1000 according to an embodiment of the present disclosure can automatically calculate multiple grading index parameter values ​​for meat pieces with high accuracy, thereby grading and rating the meat in subsequent processes.

[0056] After obtaining the trained grading index parameter model 30, the batch grading process of meat pieces can be started. That is, for a batch of meat pieces, it is only necessary to use an ordinary meat imaging detection device 10 to obtain the meat imaging information MD of each meat piece one by one. The trained grading index parameter model 30 can receive the meat imaging information MD of each meat piece and calculate and output the grading index parameter calculated value DCR' of each meat piece, or directly output the corresponding grade (i.e., the grading result). In this way, after obtaining the trained grading index parameter model 30, the grading process of batch meat pieces can be completed quickly, efficiently and accurately, increasing cost-effectiveness and improving customer satisfaction.

[0057] As an example, the grading index parameter model 30 can be a machine learning model including a neural network. By using a training model, the model can also be trained for different multiple index parameters according to different user needs, so as to ensure the output accuracy of the targeted index parameters. For example, the above purpose can be achieved by modifying the data items included in the grading truth data TD. However, the embodiments of the present disclosure are not limited to this. In this way, the meat grading system 1000 according to the embodiments of the present disclosure has higher flexibility and can be applied to more types of meat factories.

[0058] Furthermore, the training process for the grading index parameter model 30 can include training with a large amount of sample data. For example, during the training of the grading index parameter model 30, a large number of meat pieces can be used, and the ground truth data TD for each of the meat pieces can be obtained one by one to train the grading index parameter model 30 multiple times. This method can effectively improve the accuracy of the grading index parameter model 30, thereby minimizing calculation errors caused by occasional model issues and improving the reliability of model application.

[0059] As shown in Figure 3, process ①, i.e., the process of calculating the grading index parameter calculated value DCR by the grading index parameter model 30, can, as an example, include the grading index parameter model 30 performing image processing on the meat imaging information MD to segment the meat imaging information MD, and calculating the grading index parameter calculated value DCR based on the segmented meat imaging information MD. In this example, the grading index parameter model 30 can be configured to have image processing capabilities. For example, the grading index parameter model 30 can include an image processing unit. In this case, the meat imaging information MD can be directly transmitted to the grading index parameter model 30. In the grading index parameter model 30, the image processing unit first performs specific image processing on the meat imaging information MD. As mentioned above, it is necessary to segment and locate the multi-dimensional image of the meat imaging information MD. Therefore, the image processing unit can perform segmentation processing on the meat imaging information MD to segment at least one part and position information (e.g., three-dimensional coordinate information) corresponding to each part. Thus, in the subsequent calculation process, the grading index parameter model 30 can calculate the grading index parameter calculated value DCR for at least one part based on the image-processed (i.e., segmented) meat imaging information MD.

[0060] Alternatively, as another example, the grading index parameter model 30 may not include the aforementioned image processing unit, that is, it may not have image processing functionality. Instead, in process ①, the meat imaging information MD may be transmitted to an external image processing device, where it may perform specific image processing similar to that described above, thereby obtaining segmented meat imaging information MD. The grading index parameter model 30 then calculates the grading index parameter calculated value DCR based on the segmented meat imaging information MD.

[0061] FIG4 illustrates a flowchart for training a grading index parameter model 30 according to another embodiment of the present disclosure. As shown in FIG4 , the meat grading system 1000 may further include an image segmentation and positioning model 40. According to an embodiment of the present disclosure, the image segmentation and positioning model 40 may be trained using a portion of the meat imaging information MD and the ground truth grading data TD. The output of the trained image segmentation and positioning model 40 is then used to provide a portion PDCR of the calculated grading index parameter value DCR based on the meat imaging information MD.

[0062] As an example, the image segmentation and positioning model 40 can be a model for obtaining the segmentation information and position information of the parts or tissues of the meat multi-dimensional image. That is, the image segmentation and positioning model 40 can be understood as a model for performing an image processing process similar to the above on the meat imaging information MD, and thereby obtaining the part segmentation information and position information of the image. According to an embodiment of the present disclosure, the image segmentation and positioning model 40 can also be trained based on the meat imaging information MD and the graded true value data TD, and can have a training process similar to the training process of the graded index parameter model 30 as described above. The only difference between the two is that the training of the image segmentation and positioning model 40 is only based on the part of the graded true value data TD regarding the image segmentation information. In this way, a trained image segmentation and positioning model 40 can be obtained.

[0063] In this case, the trained image segmentation and positioning model 40 can obtain a calculation result based on the meat imaging information MD as the output of the trained image segmentation and positioning model 40. The output (i.e., the obtained calculation result) can be a portion PDCR of the grading index parameter calculation value DCR. As an example, the portion PDCR can include the segmentation information of at least one part of the meat piece and the position information of at least one part as described above with reference to FIG2. Then, as shown in FIG4, the trained image segmentation and positioning model 40 can output a portion PDCR of the calculated grading index parameter calculation value to the grading index parameter model 30 for use by the grading index parameter model 30.

[0064] As an example, the grading index parameter model 30 may further calculate the remaining portion of the grading index parameter calculated value DCR based on a portion of the grading index parameter calculated value PDCR and the meat imaging information MD, and then output the remaining portion of the grading index parameter calculated value DCR together with the portion of the PDCR as the grading index parameter calculated value DCR. Alternatively, as another example, the portion of the PDCR may not be used in the calculation of the remaining portion. That is, the grading index parameter model 30 may calculate the remaining portion of the grading index parameter calculated value DCR based solely on the meat imaging information MD, and then output the portion of the grading index parameter calculated value DCR together with the portion of the PDCR received from the trained image segmentation and positioning model 40 as the grading index parameter calculated value DCR.

[0065] It should be understood that the calculation of the grading index parameter calculated value DCR' for the trained grading index parameter model 30 can also include the aforementioned image processing process or the training and calculation process of the image segmentation and positioning model 40, and can be performed in a similar manner. As an example, the image processing of the meat imaging information MD can include at least one of image segmentation and positioning, image noise reduction, and image enhancement. However, the embodiments of the present disclosure are not limited to this.

[0066] Advantageously, the meat grading system 1000 described in the various embodiments of the present disclosure, through the use of a trained grading index parameter model, can automatically and accurately grade meat in batches, thereby improving meat grading efficiency, reducing customer losses, and increasing customer satisfaction. Furthermore, this system overcomes the technical issues inherent in related technologies, which rely heavily on individual worker experience, resulting in significant variability over time and thus potentially leading to grading deviations and resulting in financial losses for meat factories.

[0067] In addition, the meat grading system 1000 described in the various embodiments of the present disclosure abandons the traditional human eye observation method, so that its grading accuracy, stability and consistency will be greatly improved and enhanced, and the labor cost of hiring workers can be saved, thereby effectively improving the economic benefits of meat manufacturers.

[0068] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not a limitation herein.

[0069] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A meat grading method, the method comprising: Acquire multi-dimensional meat imaging information of meat pieces; Performing high-precision measurement on the meat pieces to obtain true value data of the grading of the meat pieces; The grading index parameter calculated values ​​are calculated based on the multi-dimensional meat imaging information, and the grading index parameter model is trained based on the grading index parameter calculated values ​​and the grading true value data to obtain a trained grading index parameter model.

2. The method according to claim 1, further comprising: The meat pieces are graded according to the grading index parameter calculation values ​​calculated by the trained grading index parameter model.

3. The method according to claim 2, wherein: The calculated value of the grading index parameter includes at least the morphological structure information and the composition information of the whole meat piece and at least one part of the meat piece, and The method further includes grading the meat pieces according to the morphological structure information and component information required for grading and referring to the grading standard.

4. The method according to claim 3, wherein: The grading index parameter calculation value includes division information of at least one part of the meat piece and position information of at least one part.

5. The method according to claim 4, wherein: The morphological structure information includes at least one of shape information, length information, width information, thickness information, and weight information of the entire meat piece and each of at least one part; The ingredient information includes at least one of the mass percentage, lean meat rate, and fat rate of the meat piece as a whole and each of at least one part.

6. The method according to claim 1, wherein: Training the grading index parameter model includes: The difference between the calculated value of the grading index parameter and the grading true value data is calculated, and the grading index parameter model is trained by comparing the difference with corresponding reference data.

7. The method according to claim 6, wherein: Training the grading index parameter model also includes: When the difference is greater than the reference data, the model parameters of the grading indicator parameter model are updated to recalculate the grading indicator parameter calculated value, and the difference between the recalculated grading indicator parameter calculated value and the grading true value data is compared with the reference data, and the calculation and comparison are repeated until the difference is less than or equal to the reference data.

8. The method according to claim 1, wherein: Calculating the calculated value of the grading index parameter includes: The grading index parameter model performs image processing on the meat imaging information to divide the meat imaging information into Information, and calculate the grading index parameter calculation value according to the divided meat imaging information.

9. The method according to claim 1, wherein: Calculating the calculated value of the grading index parameter includes: The grading index parameter model receives meat imaging information that has been divided after image processing, and calculates the grading index parameter calculation value based on the divided meat imaging information.

10. The method according to any one of claims 8 to 9, wherein: The image processing includes at least one of image segmentation and positioning processing, image noise reduction processing, and image enhancement processing.

11. The method according to claim 1, further comprising: An image segmentation and positioning model is trained using the meat imaging information and a portion of the grading truth data, and a portion of the grading index parameter calculation value is provided based on the meat imaging information using the output of the trained image segmentation and positioning model.

12. The method according to claim 1, wherein the high-precision measurement comprises: Using a high-precision imaging device to obtain true value data of the grading of the meat pieces; and / or The meat piece is divided into at least one portion and the at least one portion is measured separately to obtain true grading data of the meat piece.

13. A meat grading system, the system executing the meat grading method according to any one of claims 1 to 9, the system comprising: A meat imaging detection device, wherein the meat imaging detection device obtains multi-dimensional meat imaging information of the meat piece; a true value data acquisition device, the true value data acquisition device acquires the graded true value data of the meat pieces; A grading index parameter model is provided, wherein the grading index parameter model calculates grading index parameter calculated values ​​based on the multi-dimensional meat imaging information, and is trained based on the grading index parameter calculated values ​​and the grading true value data to obtain a trained grading index parameter model.

14. The meat grading system according to claim 13, wherein: The trained grading index parameter model grades the meat pieces according to the grading index parameter calculation values ​​calculated by the trained grading index parameter model.

15. The meat grading system of claim 13, wherein: The system also includes an image segmentation and positioning model, which is trained based on the meat imaging information and a portion of the grading true value data, and uses the trained image segmentation and positioning model to provide an output based on the meat imaging information as a part of the grading index parameter calculation value.

16. The meat grading system according to any one of claims 13 to 15, wherein: The hierarchical true value data acquisition device comprises: A manual measuring device, wherein the manual measuring device is used to cut the meat piece and directly measure at least one cut part to obtain true value data of the grade; and / or A high-precision imaging device is used to obtain high-precision imaging information of the meat pieces at a resolution higher than that of the meat imaging detection equipment, and obtain the classification true value data therefrom.

Images