摘要
本文以腊肠为研究对象,探讨了高光谱图像技术对其酸价检测的可行性。研究中,对高光谱成像系统获得的数据进行了MNF变换、PPI纯净指数计算、n-D Visualizer等处理,获得纯净的光谱数据信息。采用主成分分析,获得主成分图像,选取PC2作为分析对象,通过比较权重系数,选取六个特征波长943.28、1003.20、1136.53、1240.03、1326.95、1477.64 nm,并通过特征波长,选定1000~1500 nm波长范围作为光谱分析区域。利用PLS建模方法将高光谱数据与酸价实际值关联,获得腊肠酸价评价模型。采用一阶导+SG(17)+矢量归一化和二阶导+SG(21)+矢量归一化校正方法建立校正模型,校正集RMSECV和R^2分别为0.28,0.97和0.31和0.96,验证集RPD分别为2.92和2.89,一阶导+SG(17)+矢量归一化建立的PLS模型更适合酸价的定量检测,模型预测值平均重复性标准差为0.22,模型预测值平均相对误差为10.32%。研究结果表明,高光谱图像技术检测腊肠酸价含量是可行的。
Based on Chinese sausage,it has a discussion about the feasibility of hyperspectral imaging technique for detecting acid value.To deal with hyperspectral data by MNF（ Minimum Noise Fraction Rotation）transform, PPI （pixel purity index）index calculation, n-D Visualizer, the ＂pure＂ reflection spectrum information were collected. Principal component analysis were used to obtain the principal component images.To compare with weight coefficient of the second principal component images, the the characteristic wavelengths were 943.28,1003.20, 1136.53,1240.03, 1326,95,1477.64 rim.Choosing 1000 to 1500 nm wavelength range as the spectral analysis area.PLS modeling method was used to combine hyperspectral data with acid value practical values.The studies developed the calibration model using first derivative and SG（17）adding vector normalization,second derivative and SG（21） adding vector normalization.The RMSECV and R2 of calibration were 0.28,0.97 and 0.31,0.96 separately, and the RPD of prediction were 2.92 and 2.89 separately.Overall,the best prediction result was based on PLS established by first derivative and SG（17）adding vector for AV detection.The average of repeatability standard deviation and relative error were 0.22 and 10.32%.Therefore, hyperspectral imaging technique could be used to detect acid value on Chinese sausage.
出处
《食品工业科技》
CAS
CSCD
北大核心
2016年第5期287-291,297,共6页
Science and Technology of Food Industry
基金
国家自然科学基金资助项目(61473009)
北京市自然科学基金资助(4122020)
关键词
酸价
高光谱图像技术
PLS
主成分分析
acid value
hyperspectral imaging technique
PLS
principal component analysis
