基于机器视觉技术的肉新鲜度分级方法研究  

Research on method to freshness grading of meat based on machine vision technology

作　　者：姜沛宏[1,2] 张玉华[1,2] 钱乃余[1,2] 张长峰[1,2] 陈东杰[1] 

机构地区：[1]山东商业职业技术学院,山东省农产品贮运保鲜技术重点实验室,济南250103 [2]山东国家农产品现代物流工程技术研究中心,济南250103

出　　处：《食品科技》2015年第3期296-300,共5页Food Science and Technology

基　　金：国家科技支撑计划项目(2013BAD19B02);山东省自主创新及成果转化专项(2014ZZCX02701)

摘　　要：利用机器视觉技术对肉品新鲜度分级方法进行研究,经过图像处理提取RGB和HIS色彩模型的特征分量,分析这些特征分量在肉品贮藏期间的变化趋势,依据TVB-N含量将肉品划分为新鲜、次新鲜和腐败3个级别,应用神经网络建立基于RGBHIS特征分量的肉品新鲜度分级模型。结果显示,牛肉图像的R值随贮藏时间的延长线性降低,G、B值则随贮藏时间的延长线性增加;H值指向由红转为蓝绿色,B值随贮藏时间的延长先减后增,而I值没有明显的趋向性。运用神经网络建立肉品新鲜度分级模型判别正确率达90%以上,表明基于机器视觉技术对肉类新鲜度进行分级是可行的。Method of freshness grading of meat based on machine vision technology were studied. Through image processing to extract color features in RGB and HIS spaces, analyzing changes of meatcolor during storage. The meat could be divided into three levels of fresh, stale and putrid on the basis of TVB-N content. Using neural network to establish meat freshness graded model based on the color features. It was found that R value of beef lean tissue decreases linearly and G and B values rise linearly with storage duration extending. H value showed the beef color varies from cerise to blue-green, B value increased after first decreased and I value of beef color fluctuated in a small range. The appraising accuracies of meat freshness graded model were above 90%, the preliminary research showed that it was feasible to grade meat freshness based on machine vision technology.

关 键 词：机器视觉 色彩模型 新鲜度 神经网络 无损检测 

分 类 号：TS251.1[轻工技术与工程—农产品加工及贮藏工程]