基于高光谱技术的牛肉含水率无损检测

摘要

The moisture content of beef not only can directly affect the beef quality , but also brings great economic dam-age to the consumers , therefore this experiment provides the basis for beef quality detection by exploring the feasibility of detecting beef moisture content through hyperspectral image technology .82 samples of cows ’ back-legs are adopted as experiment materials to measure their real moisture and to collect their hyperspectral images by way of international drying method according to the specification of 5 ×4 ×1 cm.After getting the hyperspectral information of the samples , with the help of ENVI and MATLAB software , interesting areas are gained .By different pretreatment methods , artificial neural network and PLS calibration model are separately built .Though comparison of the results of these two models , PLSR cali-bration model can better predict the beef moisture content .The correlation coefficient of correction is 0 .91 the RMSEC is 0.121 the correlation coefficient of prediction set is 0.89and the RMSEP is 0.118.The results show that beef moisture content can be quickly and intact detected through hyperspectral image technology .%牛肉含水率的高低不仅直接影响牛肉品质，而且会对消费者造成经济损失。为此，通过实验探究了采用高光谱图像技术对牛肉含水率进行检测的可行性，为检测牛肉品质提供依据。采用82个牛肉后腿样本作为实验材料，按5×4×1cm的规格通过国际烘干法测量其真实含水量，并采集它们的光谱图像；获取样本的光谱信息后，通过ENVI 及MatLab 软件获取感兴趣区域。同时，利用不同的预处理方法，分别建立BP 神经网络和偏最小二乘校正模型，通过比对两种模型结果，偏最小二乘校正模型能够更有效预测牛肉含水率，校正集相关系数为0．91，校正标准差为0．121，预测集的相关系数为0．89，预测标准差为0．118。研究结果证实，利用高光谱图像技术可以快速无损检测牛肉含水率。