Method and system for non-destructive testing and grading of live pig quality

Abstract

The application discloses a kind of live pig quality nondestructive testing and grading method and system, the method includes: the two-dimensional appearance image of pig, depth image and hyperspectral imaging data are collected by inspection robot;Based on the image data collected, the live characteristics affecting pig meat quality and grade are extracted;Based on deep learning model or multivariate regression model, the mapping relationship between the live characteristics and grading standard is established, and the expected lean meat rate and carcass grade of pig are output;The grading result containing the expected lean meat rate and carcass grade and abnormal alarm picture are uploaded to the central control system in real time.The present application can ensure that grading before slaughter is carried out without interfering with the normal activities of pigs, avoiding human intervention and stress response, and improving grading efficiency and accuracy.

Description

Technical Field

[0001] This application relates to the field of livestock quality testing, and more specifically, to a method and system for non-destructive testing and grading of the quality of live pigs. Background Technology

[0002] In the livestock and poultry farming sector, traditional manual grading and slaughtering operations are not only time-consuming and labor-intensive, but also prone to triggering stress reactions during processing, affecting meat quality and production efficiency. Particularly in pig slaughtering, traditional grading methods typically rely on manual judgment and measurement by the naked eye. This process is not only labor-intensive but also susceptible to human factors, leading to inconsistent grading standards and inaccurate meat quality assessments. Furthermore, traditional methods often involve post-slaughter testing, failing to achieve real-time monitoring and evaluation of live pigs, which presents numerous challenges for farm management and decision-making. Summary of the Invention

[0003] The main purpose of this application is to provide a pre-slaughter grading scheme for livestock, which combines automated grading technology for slaughter lines with the digital management needs of pig farms. This allows farms to grasp the growth status and quality level of pigs earlier, thereby making more scientific decisions on slaughter and avoiding the errors and subjectivity caused by traditional manual grading. It can provide breeding enterprises with a more efficient and accurate operation and management method.

[0004] To achieve the above objectives, the first aspect of this application proposes a non-destructive testing and grading method for live pigs, comprising: collecting two-dimensional appearance images, depth images, and hyperspectral imaging data of pigs using an inspection robot; extracting live characteristics affecting the meat quality and grade of pigs based on the collected image data; establishing a mapping relationship between the live characteristics and grading standards based on a deep learning model or a multivariate regression model, and outputting the expected lean meat percentage and carcass grade of the pigs; and uploading the grading results containing the expected lean meat percentage and carcass grade, along with abnormal alarm images, to a central control system in real time.

[0005] In one embodiment of this disclosure, the acquisition of two-dimensional appearance images, depth images, and hyperspectral imaging data of pigs by an inspection robot includes: The inspection robot scans the pigs in the pre-slaughter waiting area along a preset path or through autonomous navigation, assigning a unique identifier to each pig and tracking and identifying each pig in real time based on a target detection algorithm. It uses a 3D binocular camera to scan each pig to obtain depth images and a high-resolution camera to obtain two-dimensional appearance images of the pigs. It uses a multispectral imager and / or a thermal imager to capture the reflectance spectral information of the pigs' skin in multiple bands.

[0006] In one embodiment of this disclosure, extracting live characteristics affecting the quality and grade of pig meat based on acquired image data includes: The acquired two-dimensional appearance images were denoised and segmented. An object detection algorithm was used to locate the pig's bounding box in the image. Based on a semantic segmentation model, the pig was segmented from the background, and key anatomical points, including the snout tip, ear base, scapular point, backfat point, lumbar vertebra point, and tail base, were extracted. The two-dimensional appearance image was fused with the depth image to generate a three-dimensional point cloud model of the pig. Based on the three-dimensional point cloud model, the pig's body length, height, volume, chest circumference, spatial position and posture, backfat thickness estimate, and rump fullness were calculated. The pig's health status and stress status were assessed based on reflectance spectral information.

[0007] In one embodiment of this disclosure, the mapping relationship between the live animal characteristics and the grading criteria is established based on a deep learning model or a multivariate regression model, and the expected lean meat percentage and carcass grade of the pigs are output, including: The collected images are labeled to record the body shape characteristics of each pig and its corresponding official grading standards. A convolutional neural network model or a multivariate regression model is trained using the labeled dataset. The trained convolutional neural network model or multivariate regression model is deployed to an embedded AI computing box as a grading model. The extracted live features are input into the deep learning model or multivariate regression model to establish a mapping relationship between the live features and the grading standards. The expected lean meat percentage and carcass grade of the pig are output. The carcass grade is a comprehensive evaluation based on lean meat percentage, fat distribution, and body shape factors, and is divided into excellent, good, medium, and poor.

[0008] In one embodiment of this disclosure, uploading the grading results, including the expected lean meat percentage and carcass grade, and abnormal alarm images to the central control system in real time includes: If the meat quality indicators of pigs are detected to be lower than the preset standard or abnormal characteristics are found, an alarm message is generated and a relevant abnormal image is attached. This information is then uploaded to the central control system in real time via a wireless network. The central control system stores, analyzes, and visualizes the data and provides feedback based on the actual meat quality results after slaughter, adjusting and optimizing the grading model accordingly.

[0009] The second aspect of this disclosure provides a non-destructive testing and grading system for live pigs, including an inspection robot and a central control system. The inspection robot and the central control system are wirelessly connected. The inspection robot is used to collect two-dimensional appearance images, depth images, and hyperspectral imaging data of pigs. Based on the collected image data, it extracts live characteristics that affect the meat quality and grade of the pigs. Based on a deep learning model or a multivariate regression model, it establishes a mapping relationship between the live characteristics and grading standards, and outputs the expected lean meat percentage and carcass grade of the pigs. The grading results, including the expected lean meat percentage and carcass grade, and abnormal alarm images are uploaded to the central control system in real time. The central control system is used to store, analyze, and visualize the data, and provide feedback based on the actual meat quality results after slaughter, adjusting and optimizing the grading model.

[0010] In one embodiment of this disclosure, the inspection robot integrates a lidar and an inertial measurement unit, uses the SLAM algorithm for environmental modeling and localization, autonomously plans its path within the slaughter pen, avoids obstacles, and maintains the optimal observation distance to the pigs.

[0011] In one embodiment of this disclosure, the inspection robot integrates a high-resolution camera, a 3D binocular camera, a multispectral imager, and / or a thermal imager. It is used to acquire two-dimensional appearance images of pigs through the high-resolution camera, obtain three-dimensional contour data of pigs through the 3D binocular camera, capture the reflectance spectrum information of pigs' skin in multiple bands through the multispectral imager, and / or detect the temperature distribution on the surface of pigs' bodies through the thermal imager to help determine whether pigs have a stress response. The inspection robot integrates an embedded AI computing box to denoise and segment the collected two-dimensional appearance images, extracting key anatomical points including the nose tip, ear base, scapula point, backfat point, lumbar vertebra point, and tail base. It then fuses the two-dimensional appearance images with depth images to generate a three-dimensional point cloud model of the pig. Based on this model, it calculates the pig's body length, height, volume, chest circumference, spatial position and posture, estimated backfat thickness, and rump fullness, assessing the pig's health and stress status. Using deep learning or regression models, it infers the relationship between pig characteristics and grading standards, predicting the lean meat percentage and carcass grade after slaughter, and uploading the grading results and abnormal alarm images to the central control system.

[0012] A third aspect of this disclosure provides an electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to cause the at least one processor to perform the live pig quality non-destructive testing and grading method provided in the first aspect.

[0013] The fourth aspect of this disclosure provides a computer-readable storage medium storing computer instructions for causing a computer to perform the non-destructive testing and grading method for live pig quality provided in the first aspect.

[0014] The technical solutions provided by the embodiments of this application may include the following beneficial effects: By collecting appearance and depth data of pigs and using deep learning models for data analysis and processing, it is possible to automatically detect the size and health status of pigs, thereby enabling accurate pre-slaughter grading. This avoids human error and animal stress reactions in traditional methods. Combined with AI algorithms, it can predict the growth process and final slaughter grade of pigs, allowing for advance planning of breeding and slaughter strategies, thus improving the efficiency and accuracy of pre-slaughter grading of pigs. Attached Figure Description

[0015] The accompanying drawings, which form part of this application, are used to provide a further understanding of the application and to make other features, objects, and advantages of the application more apparent. The illustrative embodiments and descriptions of this application are used to explain the application and do not constitute an undue limitation of the application. In the drawings: Figure 1 A schematic diagram illustrating the non-destructive testing and grading method for live pig quality provided in this application; Figure 2 A flowchart of the live pig quality non-destructive testing and grading system provided in this application; Figure 3 This is a block diagram of an electronic device provided in an embodiment of the present disclosure. Detailed Implementation

[0016] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0017] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0018] To avoid causing stress to pigs due to human interference or contact, which could affect meat quality, pre-slaughter grading should be conducted under non-contact, low-stress conditions. This disclosure effectively combines automated grading technology for slaughter lines with the digital management needs of pig farms. Through pre-slaughter grading, farms can grasp the growth status and quality level of pigs earlier, thus making more scientific decisions regarding slaughter. This pre-prediction avoids the errors and subjectivity of traditional manual grading and aligns with the precise trading model of "high quality, high price," providing pig farming enterprises with a more efficient and accurate operational management method.

[0019] The implementation of this disclosure relies on precise monitoring and data analysis of live pigs. By collecting real-time data through sensors and combining it with AI algorithms, the growth process and final slaughter grade of pigs can be predicted, and breeding plans and slaughter strategies can be arranged in advance to improve the efficiency and accuracy of pre-slaughter grading of pigs.

[0020] Figure 1 A flowchart illustrating the non-destructive testing and grading method for live pig quality provided in this application. (Refer to...) Figure 1 As shown, in step S102, the inspection robot first collects two-dimensional appearance images, depth images and hyperspectral imaging data of the pigs.

[0021] The inspection robot scans the pigs in the pre-slaughter pen along a preset path or autonomously navigates, assigning a unique identifier to each pig and tracking and identifying each pig in real time based on a target detection algorithm. Specifically, it uses LiDAR or other sensors to create a 3D model of the pre-slaughter pen, identifying elements such as obstacles, pigs, and fences. Path planning algorithms (such as A / B) are then used to... (e.g., Dijkstra) generates an optimal inspection path for the inspection robot, avoiding collisions with pigs and other obstacles. The robot can adjust its path in real time during the inspection to adapt to the dynamic environment and ensure the completion of comprehensive monitoring tasks.

[0022] The inspection robot assigns a unique identifier to each pig, which is achieved through the following methods: RFID tags: Pigs are tagged with RFID tags, and the inspection robot uses an RFID reader to scan and identify each pig, thereby obtaining its identification information. Computer vision: Object detection algorithms (such as YOLO, Faster R-CNN, etc.) are used to identify and track the location and status of each pig in real time. Through deep learning models, the system can quickly process image data, ensuring accurate identification of each pig even in complex environments.

[0023] Each pig was scanned using a 3D binocular camera to acquire depth images, and a high-resolution camera was used to acquire two-dimensional images of the pig's appearance. Multispectral imagers and / or thermal imagers were used to capture the reflectance spectral information of the pig's skin in multiple bands.

[0024] Subsequently, in step S104, live characteristics affecting the meat quality and grade of pigs are extracted based on the acquired image data.

[0025] First, the acquired two-dimensional appearance images are denoised. Object detection algorithms (such as YOLO and Faster R-CNN) are used to locate the pig's bounding box in the image. Based on semantic segmentation models (such as U-Net and DeepLab), the pig is segmented from the background, and key anatomical points, including the snout tip, ear base, scapular point, backfat point, lumbar vertebrae, and tail base, are extracted. The two-dimensional appearance images are then fused with depth images to generate a three-dimensional point cloud model of the pig. Based on this model, the pig's body length, height, volume, chest circumference, spatial position and posture, estimated backfat thickness, and rump fullness are calculated. The pig's health and stress status are assessed based on the reflectance spectral information. For example, changes in reflectance in specific bands may be related to inflammation, fatigue, or other health problems in the pig.

[0026] In step S106, a mapping relationship between the live animal characteristics and the grading criteria is established based on a deep learning model or a multivariate regression model, and the expected lean meat percentage and carcass grade of the pigs are output.

[0027] According to one embodiment of this disclosure, the collected images are labeled to record the body shape characteristics (such as body length, height, chest circumference, etc.) of each pig and the corresponding official grading standards (such as lean meat percentage and carcass grade). A suitable convolutional neural network structure (such as ResNet, VGG, EfficientNet, etc.) is selected for image processing. If a multivariate regression model is selected, a suitable regression equation needs to be constructed, considering the relationship between multiple input features (such as body shape parameters) and output results (lean meat percentage, carcass grade). The labeled dataset is used to train the convolutional neural network model or multivariate regression model, and the trained convolutional neural network model or multivariate regression model is deployed to an embedded AI computing box as a grading model.

[0028] The extracted live features are input into the deep learning model or multivariate regression model to establish a mapping relationship between the live features and the grading standards, and output the expected lean meat percentage and carcass grade of the pig. The carcass grade is a grade comprehensively evaluated based on lean meat percentage, fat distribution and body shape factors. Usually, the carcass grade of pigs is divided into several grades, such as excellent, good, medium and poor.

[0029] Finally, in step S108, the grading results, including the expected lean meat percentage and carcass grade, and abnormal alarm images are uploaded to the central control system in real time.

[0030] In one embodiment of this disclosure, an embedded AI computing box continuously monitors the live characteristics of pigs and calculates their expected lean meat percentage and carcass grade using a pre-trained model. A threshold standard is set; when the monitored meat quality indicators (such as lean meat percentage, fat thickness, etc.) of the pigs are found to be below the preset standard, the system identifies it as an anomaly. Simultaneously, abnormal features (such as body deformities or other meat quality defects) in the images are detected. Once an anomaly is detected, an alarm message is immediately generated, recording the anomaly type, time, location, and other relevant information, along with an image of the anomaly for subsequent analysis. The alarm message is formatted, including: pig ID, expected lean meat percentage and carcass grade, description of the abnormal feature, anomaly image (JPEG / PNG format), timestamp, etc., and the above information is sent to the central control system in real time using an appropriate protocol (such as HTTP / HTTPS).

[0031] In the central control system, a database (such as MySQL, MongoDB, etc.) is established to store received alarm information and grading results. Data analysis tools are used to organize and analyze the uploaded data, identifying abnormal trends and patterns for deeper analysis. Visualization tools are used to display the data in chart form, intuitively presenting changes in various indicators and the distribution of abnormal events, helping managers quickly understand the overall health status of the pigs.

[0032] After slaughter, the actual meat quality test results are compared with the prediction results in the system to collect data and evaluate the model's accuracy. Based on the feedback information from the post-slaughter meat quality results, the bias in the model's predictions is analyzed, and features or algorithms that need optimization are identified. The grading model is retrained or adjusted to improve future prediction accuracy. By continuously accumulating new data and feedback, the model's performance and system effectiveness are optimized.

[0033] As can be seen from the above scheme, this disclosure, by collecting the appearance and depth data of pigs and using deep learning models for data analysis and processing, can achieve automated detection of pig body size and health status, thereby enabling accurate pre-slaughter grading. This avoids human error and animal stress reactions in traditional methods. Combined with AI algorithms, it can predict the growth process and final slaughter grade of pigs, allowing for advance planning of breeding and slaughter strategies, thus improving the efficiency and accuracy of pre-slaughter grading of pigs.

[0034] This disclosure also provides a non-destructive testing and grading system for live pigs, comprising: an inspection robot and a central control system, wherein the inspection robot and the central control system are wirelessly connected; the inspection robot is used to collect two-dimensional appearance images, depth images, and hyperspectral imaging data of pigs; extract live characteristics affecting the meat quality and grade of pigs based on the collected image data; establish a mapping relationship between the live characteristics and grading standards based on a deep learning model or a multivariate regression model, and output the expected lean meat percentage and carcass grade of the pigs; upload the grading results containing the expected lean meat percentage and carcass grade and abnormal alarm images to the central control system in real time; the central control system is used to store, analyze, and visualize the data, and provide feedback based on the actual meat quality results after slaughter, adjusting and optimizing the grading model.

[0035] Among them, the inspection robot integrates lidar and inertial measurement unit, uses SLAM algorithm for environmental modeling and localization, autonomously plans its path in the slaughter pen, avoids obstacles and maintains the best observation distance with the pigs.

[0036] The inspection robot integrates a high-resolution camera, a 3D binocular camera, a multispectral imager, and / or a thermal imager. It is used to acquire two-dimensional appearance images of pigs through the high-resolution camera, obtain three-dimensional contour data of pigs through the 3D binocular camera, capture the reflectance spectrum information of pigs' skin in multiple bands through the multispectral imager, and / or detect the temperature distribution of pigs' body surface through the thermal imager to help determine whether pigs have a stress response. The inspection robot integrates an embedded AI computing box to denoise and segment the collected two-dimensional appearance images, extracting key anatomical points including the nose tip, ear base, scapula point, backfat point, lumbar vertebra point, and tail base. It then fuses the two-dimensional appearance images with depth images to generate a three-dimensional point cloud model of the pig. Based on the three-dimensional point cloud model, it calculates the pig's body length, height, volume, chest circumference, spatial position and posture, backfat thickness estimate, and rump fullness to assess the pig's health and stress status. Based on deep learning or regression models, it infers the relationship between pig characteristics and grading standards, predicts the lean meat percentage and carcass grade of the pig after slaughter, and uploads the grading results and abnormal alarm images to the central control system.

[0037] The system can be divided into a perception layer, an algorithm layer, and an application layer. The perception layer is mainly responsible for image acquisition and health status assessment of pigs. It collects multi-dimensional data through technologies such as machine vision, 3D point cloud reconstruction, and multispectral / hyperspectral imaging, and provides a foundation for subsequent data processing.

[0038] In one embodiment of this disclosure, the robot inspects along a preset path within the narrow passage of the pre-slaughter pen. It navigates autonomously using a SLAM (Simultaneous Localization and Mapping) algorithm, and employs a combination of lidar and IMU (Inertial Measurement Unit) for autonomous navigation, maintaining a fixed distance from the pigpen to avoid interfering with its normal activities. Upon detecting a pig, the robot briefly stops or decelerates to perform key point detection and 3D point cloud reconstruction. Using the 3D point cloud model, the system can accurately extract features such as the pig's body shape and backfat, providing support for grading.

[0039] Each pig in the pen is scanned in real time using an RGB-D binocular camera or a 3D ToF camera, capturing color images and depth information in real time. This avoids interference from changes in coat color and lighting, allowing for accurate calculation of body shape data such as body length, height, and chest circumference. High-resolution cameras are used to capture images of the pig's body shape, backfat thickness, and rump fullness, providing necessary input data for deep learning models. Multispectral / hyperspectral imaging is used to assess the pig's skin health (e.g., erythema, bruising) and the presence of stress, factors that affect meat quality and serve as auxiliary deductions during grading.

[0040] The algorithm layer is responsible for processing, analyzing, and modeling the data collected by the perception layer. Its core function is to use deep learning and regression analysis methods, combined with the pig's body shape characteristics and meat quality standards, to make accurate grading predictions.

[0041] The inspection robot uses object detection models such as YOLOv8 or Faster R-CNN to identify each pig in the scene in real time. To avoid occlusion and overlap of pigs, SORT or DeepSORT algorithms are used to assign a unique ID to each pig, ensuring accurate tracking and measurement. Image denoising and segmentation algorithms are used to separate the pigs from the complex background, ensuring data accuracy and reliability. Multiple key anatomical sites of the pigs (such as the snout tip, backfat point, and tail root) are identified through algorithms. Two-dimensional images are fused with depth maps to generate a three-dimensional point cloud model of the pig, which is then analyzed and its volume is calculated.

[0042] A convolutional neural network (CNN) model is trained to automatically recognize and learn the body shape characteristics of pigs and map them to official grading standards. By learning features such as backfat thickness and rump fullness, the model predicts the lean meat percentage and carcass grade of pigs. The extracted features (such as backfat thickness and eye muscle area) are substituted into the regression equation to predict the lean meat percentage and carcass grade of pigs after slaughter, achieving accurate grading.

[0043] The application layer primarily relies on intelligent robots for autonomous navigation and positioning, ensuring stable system operation and real-time response. Through edge computing units, the system can perform real-time image processing on the robot, reducing latency and improving classification efficiency. The classification results are then uploaded to the central control system in real-time via 5G / 4G or Wi-Fi 6 technology. The edge computing units are responsible for processing image data on the robot in real-time, reducing data transmission latency and enabling "inspection and classification simultaneously." This approach not only improves data processing efficiency but also reduces reliance on cloud computing resources, achieving more flexible real-time response.

[0044] As can be seen from the above scheme, this disclosure utilizes machine vision and deep learning technologies to ensure pre-slaughter grading without interfering with the normal activities of pigs, avoiding human intervention and stress responses in traditional methods. By combining convolutional neural networks and regression models, the grading system can predict the lean meat percentage and carcass grade of pigs based on visual data. Data processing is performed on the robot side, and monitoring is conducted by uploading grading results and abnormal images in real time, reducing data transmission latency and improving grading efficiency and accuracy.

[0045] This disclosure also provides an electronic device, such as... Figure 3 As shown, the electronic device includes one or more processors 301 and a memory 302. Figure 3 Take processor 301 as an example.

[0046] The electronic device may also include an input device 303 and an output device 304.

[0047] The processor 301, memory 302, input device 303, and output device 304 can be connected via a bus or other means. Figure 3 Taking the example of a connection between China and Israel via a bus.

[0048] Processor 301 can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or combinations of the above types of chips. The general-purpose processor can be a microprocessor or any conventional processor.

[0049] The memory 302, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the methods in the embodiments of this disclosure. The processor 301 executes various functional applications and data processing of the server by running the non-transitory software programs, instructions, and modules stored in the memory 302, thereby implementing the methods in the above-described method embodiments.

[0050] Memory 302 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created by the use of the processing device operated by the server. Furthermore, memory 302 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory 302 may optionally include memory remotely located relative to processor 301, and these remote memories can be connected to a network connection device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0051] One or more modules are stored in memory 302 and, when executed by one or more processors 301, perform the methods shown in the above embodiments.

[0052] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0053] Obviously, those skilled in the art should understand that the various units or steps of this application described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device, or fabricating them separately as individual integrated circuit modules, or fabricating multiple modules or steps into a single integrated circuit module. Thus, this application is not limited to any particular combination of hardware and software.

[0054] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A non-destructive testing and grading method for the quality of live pigs, characterized in that, include: Two-dimensional appearance images, depth images, and hyperspectral imaging data of pigs are collected by inspection robots; Based on the collected image data, live characteristics that affect the meat quality and grade of pigs are extracted; The mapping relationship between the live animal characteristics and the grading criteria is established based on a deep learning model or a multivariate regression model, and the expected lean meat percentage and carcass grade of the pigs are output. The grading results, including the expected lean meat percentage and carcass grade, along with abnormal alarm images, are uploaded to the central control system in real time.

2. The method for non-destructive testing and grading of live pig quality according to claim 1, characterized in that, The collection of two-dimensional appearance images, depth images, and hyperspectral imaging data of pigs by the inspection robot includes: The inspection robot scans the pigs in the pigpen along a preset path or by autonomous navigation in the pre-slaughter waiting area, assigns a unique identification to each pig, and tracks and identifies each pig in real time based on a target detection algorithm. Each pig was scanned using a 3D binocular camera to obtain a depth image of the pig, and a high-resolution camera was used to obtain a two-dimensional appearance image of the pig. Use multispectral imagers and / or thermal imagers to capture the reflectance spectra of pig skin in multiple bands.

3. The method for non-destructive testing and grading of live pig quality according to claim 2, characterized in that, The extraction of live characteristics affecting the quality and grade of pig meat based on the acquired image data includes: The acquired two-dimensional appearance image is denoised, the pig's bounding box is located in the image using a target detection algorithm, and the pig is segmented from the background based on a semantic segmentation model. Key anatomical points, including the tip of the nose, the base of the ears, the scapula, the backfat, the lumbar vertebrae, and the base of the tail, are extracted. The two-dimensional appearance image and the depth image are fused to generate a three-dimensional point cloud model of the pig. Based on the three-dimensional point cloud model, the pig's body length, body height, volume, chest circumference, spatial position and posture, backfat thickness estimate, and rump fullness are calculated. The health and stress status of pigs are assessed based on the reflected light spectrum information.

4. The method for non-destructive testing and grading of live pig quality according to claim 2, characterized in that, The mapping relationship between the live animal characteristics and the grading criteria is established based on a deep learning model or multivariate regression model, and the expected lean meat percentage and carcass grade of the pigs are output, including: The collected images are labeled to record the body shape characteristics of each pig and its corresponding official grading standards. The labeled dataset is used to train a convolutional neural network model or a multivariate regression model. The trained convolutional neural network model or multivariate regression model is then deployed to an embedded AI computing box as a grading model. The extracted live features are input into the deep learning model or multivariate regression model to establish a mapping relationship between the live features and the grading standards, and output the expected lean meat percentage and carcass grade of the pigs. The carcass grade is a grade comprehensively evaluated based on lean meat percentage, fat distribution and body shape factors, and is divided into excellent, good, medium and poor.

5. The method for non-destructive testing and grading of live pig quality according to claim 4, characterized in that, The real-time uploading of the grading results, including the expected lean meat percentage and carcass grade, and abnormal alarm images to the central control system includes: If the meat quality indicators of pigs are detected to be lower than the preset standard or abnormal characteristics are detected, an alarm message is generated and a relevant abnormal image is attached, which is then uploaded to the central control system in real time via wireless network. The central control system stores, analyzes, and visualizes data, and provides feedback based on the actual meat quality results after slaughter, adjusting and optimizing the grading model.

6. A non-destructive testing and grading system for the quality of live pigs, characterized in that, It includes an inspection robot and a central control system, which are wirelessly connected. The inspection robot is used to collect two-dimensional appearance images, depth images, and hyperspectral imaging data of pigs; and to extract live characteristics that affect the meat quality and grade of pigs based on the collected image data. The mapping relationship between the live animal characteristics and the grading standards is established based on a deep learning model or a multivariate regression model, and the expected lean meat percentage and carcass grade of the pigs are output. The grading results containing the expected lean meat percentage and carcass grade and abnormal alarm images are uploaded to the central control system in real time. The central control system is used to store, analyze and visualize the data, and provide feedback based on the actual meat quality results after slaughter to adjust and optimize the grading model.

7. The live pig quality non-destructive testing and grading system according to claim 6, characterized in that, The inspection robot integrates lidar and inertial measurement unit, uses SLAM algorithm for environmental modeling and localization, autonomously plans its path within the slaughter pen, avoids obstacles, and maintains the optimal observation distance to the pigs.

8. The method for non-destructive testing and grading of live pig quality according to claim 6, characterized in that, The inspection robot integrates a high-resolution camera, a 3D binocular camera, a multispectral imager, and / or a thermal imager. It is used to acquire two-dimensional appearance images of pigs through the high-resolution camera, obtain three-dimensional contour data of pigs through the 3D binocular camera, capture the reflectance spectrum information of pigs' skin in multiple bands through the multispectral imager, and / or detect the temperature distribution on the surface of pigs' bodies through the thermal imager to help determine whether pigs have a stress response. The inspection robot integrates an embedded AI computing box to denoise and segment the collected two-dimensional appearance images, extracting key anatomical points including the nose tip, ear base, scapula point, backfat point, lumbar vertebra point, and tail base. It then fuses the two-dimensional appearance images with depth images to generate a three-dimensional point cloud model of the pig. Based on this model, it calculates the pig's body length, height, volume, chest circumference, spatial position and posture, estimated backfat thickness, and rump fullness, assessing the pig's health and stress status. Using deep learning or regression models, it infers the relationship between pig characteristics and grading standards, predicting the lean meat percentage and carcass grade after slaughter, and uploading the grading results and abnormal alarm images to the central control system.

9. An electronic device, characterized in that, The electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to cause the at least one processor to perform the non-destructive testing and grading method for live pig quality as described in any one of claims 1-5.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the non-destructive testing and grading method for live pig quality as described in any one of claims 1-5.

Images