不同标准板下猪肉肉色评分模型构建及其测定精度比较

Development of Pork Color Scoring Models Based on Different Standards and Comparative Analysis of Their Measurement Accuracy

摘要:

针对目前猪肉肉色人工主观评分效率低、主观性强、易疲劳，以及基于不同肉色标准板的客观评分一致性和可比性差等问题，提出了一种基于肉色标准评分板图像的猪肉肉色评分模型快速构建方法，并系统开展了基于不同肉色标准评分板图像的猪肉肉色评分模型的测定精度比较。首先采用扫描仪采集了两种猪肉肉色标准评分板和540头试验猪肉样图像，开展了基于两种猪肉肉色标准评分板人工主观评分与基于欧氏距离肉色客观评分比较分析；然后采用Matlab软件提取了猪肉肉色标准评分板图像的L、a、b值，并采用岭回归方法构建了肉色标准评分板图像L、a、b值与其肉色等级的岭回归方程，利用试验猪肉样图像对岭回归方程截距进行了校准，获得了猪肉肉色评分模型；最后对基于不同肉色标准评分板图像的猪肉肉色评分模型在540头试验猪肉样中肉色评分一致性与精度进行评价。试验结果表明，以两种肉色标准评分板图像的参数L为基础构建的肉色评分模型在评分一致性方面有更佳表现，在540头试验猪肉样中评分一致性指数超过92%，评分精度超过91%；而基于两种肉色标准评分板图像的参数L、a构建的评分模型则展现出更佳的评分精度，在540头试验猪肉样中评分精度超过95%，但评分一致性指数仅65%左右。本文提出的猪肉肉色评分模型测定精度高、模型参数校准快、不同标准评分板的测定精度一致性高，这为精准、快速、自动化猪肉肉色评分装备研制提供了理论和技术支撑。

Abstract:

Aiming to address the issues of low efficiency, high subjectivity, and fatigue associated with artificial scoring of pork color, as well as the poor consistency and comparability of objective scoring based on different pork color standard boards, a rapid construction method for pork color scoring models was proposed based on images of pork color standard boards. A systematic comparison of the scoring accuracy of pork color models based on different standard boards was conducted. Firstly, images of two pork color standard boards and 540 pork samples were collected by using a scanner, followed by a comparative analysis of subjective artificial scoring and objective scoring based on Euclidean distance for pork color. The L，a，b values of the pork color standard board images were extracted by using Matlab. Subsequently, ridge regression was applied to establish the ridge regression equations between the L，a，b values of the standard board images and their respective color grades. The intercepts of the equations were calibrated by using the pork sample images to obtain the pork color scoring models. The consistency and accuracy of the pork color scoring models based on different standard board images were evaluated by using the 540 pork samples. The results demonstrated that the scoring models based on the L parameter of the two standard panels achieved optimal consistency, with a consistency rate exceeding 92% and an accuracy rate exceeding 91%. In contrast, the scoring models based on the L，a parameters exhibited the highest scoring accuracy, achieving an accuracy rate exceeding 95%, but with a consistency rate only around 65%. The proposed pork color scoring models demonstrated high measurement precision, rapid parameter calibration, and strong consistency in measurement accuracy across different standard boards. The research result can provide theoretical and technical support for the development of precise, rapid, and automated pork color scoring systems.