可见-近红外光谱成像法快速检测猪肉新鲜度.doc

文档介绍：
可见-近红外光谱成像法快速检测猪肉新鲜
度#
汪希伟1,赵茂程1,居荣华2,陈湘来3**
5
10
15
20
25
30
35
40
(1. 南京林业大学机械电子工程学院,南京 210037;
2. 南京食品包装机械研究所,南京 210029;
3. 南京雨润肉品有限公司,南京 210041)
摘要:为实现鲜猪肉新鲜度快速无损检测,从可见-近红外光谱图像提取针对鲜猪肉挥发性
盐基氮(TVB-N)含量的特征波段。原始光谱为经反射率校正后的光谱图像中肌肉区域的均
值光谱。经标准偏差归一化方法(SNV)预处理后,建立光谱和 TVB-N 含量的偏最小二乘
回归(PLSR)模型。提出通过加权回归系数峰值波段组合建模性能分析算法,提取最佳特
征波长组。试验结果表明,基于 570,595,610,700,760,880 及 910 nm 特征波长组建立
的偏最小二乘回归模型对 TVB-N 含量预测效果最佳,“留一法”交互验证确定系数(R2cv)
与均方根误差(RMSEcv)分别为 和 mg/100g。该特征波长提取方法尤其适用于电
调制滤镜类成像光谱采集系统的高效特征光谱提取。
关键词:特征波长提取;光谱降维;偏最小二乘回归;挥发性盐基氮;猪肉
中图分类号:
Rapid Estimation of Pork Freshness using Visible and
Near-infrared Spectral Imaging
WANG Xiwei1, ZHAO Maocheng1, JU Ronghua2, CHEN Xianglai3
(1. Mechanical and Electronic Engineering School, Nanjing Forestry University, Nanjing 210037;
2. Nanjing Institute of Food Packaging Machinery;
3. Nanjing Yurun Meat Co. Ltd, Nanjing 210041)
Abstract: In order to realize the rapid and non-destructive evaluation of the freshness of pork,
feature wavelengths for total volatile basic nitrogen (TVB-N) were selected from the visible/near
infrared spectral images. The raw spectra were generated from the mean spectra of the muscle
region from the reflectance calibrated spectral images. After the standard normal variate (SNV)
preprocess, the spectra coupled with TVB-N references were subjected to a partial least square