基于高光谱技术的灵武长枣糖度预测模型研究

SUGAR CONTENT PREDICTION MODEL OF LINGWU LONG JUJUBE BY HYPERSPECTRAL IMAGING TECHNIQUE

灵武长枣糖度是反映其品质的重要指标之一.利用高光谱成像技术对灵武长枣的糖度进行无损检测研究.采用多元散射校正、标准正态变量变换和Savitzky-Golay平滑对900-1 700 nm及400-1 000 nm波段范围内的原始光谱进行预处理,选取最优的预处理方法,最后分别建立灵武长枣糖度的PCA和PLSR预测模型,优选最佳模型.结果表明,400-1 000 nm光谱数据经过多元散射校正后的光谱建立的预测模型效果较好,900-1 700 nm光谱数据则是经过Savitzky-Golay平滑后的光谱建立的预测模型较好;两者均是采用PLSR模型建模效果较好,其校正模型和验证模型的相关系数分别为0.938、0.916和0.823、0.864.研究表明,采用高光谱成像技术对灵武长枣糖度的无损检测是可行的.

Sugar content is one of the major indicators representing the quality of Lingwu long jujube. In this paper,a hyperspectral imaging technique was used for non-destructively detecting the sugar content in Lingwu long jujube. Multiplicative scatter correction, standard normal variable transformation and Savitzky-Golay smoothing were used to preprocess the original spectra in the range of 400 to 1 000 nm and 900 to 1 700 nm to select the optimum preprocessing method;and finally,sugar content prediction models based on PCA（principal component analysis） and PLSR（partial least squares regression） were constructed respectively to select the optimum model. The results showed that the prediction model constructed based on the 400 to 1 000 nm spectra data processed by multiplicative scatter correction had a good effect;the prediction model constructed based on the 900 to 1700 nm spectra data processed by Savitzky-Golay smoothing had a good effect; the models constructed by PLSR had a good effect;and the correlation coefficient of the correction model and the validation model of the above two models were respectively 0.938,0.916,and 0.823 and 0.864. The study shows that hyperspectral imaging technique is feasible for nondestructive detection of sugar content in Lingwu long jujube.