摘要
建立了牛肉基于TVB-N、菌落总数、pH值和肉色参数L*多个指标的储存期预测模型,利用可见近红外光谱(Vis/NIR)技术结合区间偏最小二乘(iPLS)和遗传算法(GA)建立了各个指标的PLS预测模型,实现了多指标综合无损快速预测4℃下牛肉的储存期。用iPLS和iPLS-GA提取有效波长变量建立PLS预测模型,以预测相关系数和预测标准差作为模型评价标准,结果表明用iPLS-GA选择变量建立的各个指标的PLS预测模型均优于全波段和iPLS组合的PLS模型。由多个指标的预测值和储存期的预测模型,对校正集和预测集样品储存期进行预测,其预测相关系数和标准差分别是0.903,0.897和1.88,2.24。说明利用光谱技术结合得出的储存期预测模型可以实现多指标综合预测牛肉储存期,为无损快速检测牛肉储存期或货架期提供了一种新方法。
The prediction model of beef＇s storage time was established based on multi indexes of fresh beef, such as TVB-N, colony total, pH value, and L＊ parameter. Visible and near-infrared spectroscopy （Vis/NIR） combined with interval PLS （iPLS） and genetic algorithm（GA） was investigated for establishing PLS calibration model of above 4 indexes, respectively, and rapid and nondestructive prediction of the storage time of fresh beef stored at 4 ℃ was realized. PLS models of 4 indexes were built with full spectrum and effective variables selected by iPLS and iPLS-GA method, respectively. The performance of each model was evaluated according to two correlations coefficients（R） and standard error （SE） of calibration and prediction sets. Ex- perimental results showed that the performance of all models built with effective variable selected by iPLS-GA was better than full spectrum and iPLS. The storage time of calibration and prediction sets of beef samples was predicted by storage time model with predicted values of above 4 indexes, and was achieved as follows： Re=0. 903, Rp =0. 897, SEC=1. 88 and SEP=2. 24. The study demonstrated that the beef＇s storage time can be synthetically predicted with multi-index by using visible and near-in- frared spectroscopy combined with the prediction model of beef＇s storage time. This provides a new method for rapid and non- destructive detection of beef＇s storage time or shelf life.
出处
《光谱学与光谱分析》
SCIE
EI
CAS
CSCD
北大核心
2012年第12期3242-3246,共5页
Spectroscopy and Spectral Analysis
基金
国家自然科学基金项目(30771244)
公益性行业(农业)科研经费项目(201003008)资助
关键词
可见近红外光谱
牛肉储存期
多指标检测
变量选择
偏最小二乘
Visible and near-infrared spectroscopy
Beef storage time
Multiple determination
Variable selection
PLS
