To ensure that pork quality inspection processes meet the line speed requirements of the meat processing industry, and to improve the accuracy, stability, and efficiency of quality grading for pork tenderloin, this study developed an industrial online grading system for pork loin quality based on machine vision and image processing technology. Key components of the system hardware were designed and selected, then a machine vision software system was developed using NI-VBAI software as the core processor. Finally, bench performance and industrial application tests were conducted. In the performance test, the grading system achieved optimal performance at a sample conveyor speed of 0.25 m/s, adjacent sample spacing of 15 mm, and a region of interest of 300 cm². Under optimal conditions, the system exhibited average accuracies of 81.55% for color grading and 70.83% for marbling grading. Further, regarding color and marbling grading, the errors between industrial application tests and performance tests were 0.76% and 0.29%, respectively, the mean running response times of the system were 0.86 s and 1.08 s, respectively, and the coefficients of variation were 7.92% and 9.36%, respectively. Overall, the system can ensure a grading accuracy and speed that satisfies the operational requirements of processing production lines.

为了确保猪肉质量检查过程满足肉类加工行业的线速度要求，并提高猪里脊肉质量分级的准确性，稳定性和效率，本研究开发了一种基于机器的猪里脊肉质量在线分级系统视觉和图像处理技术。设计并选择了系统硬件的关键组件，然后以NI-VBAI软件为核心处理器开发了机器视觉软件系统。最后，进行了台式性能和工业应用测试。在性能测试中，分级系统在样品传送带速度为0.25 m / s，相邻样品间距为15 mm，目标区域为300 cm ^{2的情况下}获得了最佳性能。。在最佳条件下，该系统显示出的平均色彩分级准确度为81.55％，大理石花纹分级平均准确度为70.83％。此外，关于颜色和大理石花纹的等级，工业应用测试和性能测试之间的误差分别为0.76％和0.29％，系统的平均运行响应时间分别为0.86 s和1.08 s，变异系数为7.92 ％和9.36％。总体而言，该系统可以确保满足加工生产线操作要求的分级精度和速度。