中文摘要：

基于近红外漫反射光谱进行了马铃薯淀粉含量快速检测,并分析了表皮对光谱及淀粉含量检测精度的影响问题。本文对110个样品的漫反射光谱进行平均化(Mean centering)、微分处理及Norris滤波,建模方法选用主成分回归(PCR)及偏最小二乘法(PLS)。检测结果表明:近红外漫反射光谱检测马铃薯淀粉含量具有可行性,但表皮对光谱及检测精度有影响。采用PLS法对经Norris滤波处理的去皮马铃薯一阶微分光谱与淀粉含量建模,效果最好,相关系数r为0.893,根校正偏差RMSEC为1.01%;对预测集样品预测,均方根预测偏差RMSEP为1.38%,精度明显优于带皮马铃薯的建模及预测精度(r为0.834,RMSEC为1.29%,RMSEP为1.74%)。果蔬品质光谱检测中,如要实现样品品质精确检测,表皮影响需进一步校正。

英文摘要：

In this paper , nondestructive and quick determination of starch content of potato was studied using NIR diffuse reflectance spectroscopy , and the influence of peel on spectra and detecting precision was analyzed .The mean centering method , derivative , norris derivative filtering methods were used to processes the spectra , and the partial least square method ( PLS) and principal component regression method ( PCR) were used to develop calibration models .110 samples detecting results showed that it is feasible to detect potato starch content using NIR diffuse reflectance spectroscopy , and the peel influence the spectra and measuring precision .The performance of the model for starch content of peeled potato using the first Norris derivative spectra combined with PLS method ( r=0 .893 , RMSEC=1 .01%, RMSEP=1 .38%) , was more better than that for intact potato starch ( r=0 .834 , RMSEC=1 .29%, RMSEP=1 .74%) .The influence of peel should be corrected in fruit and vegetable quality detecting using NIR spectroscopy .