鲜枣可溶性固形物最小二乘支持向量机动态检测研究

Dynamic Detection for Soluble Solids Content of Fresh Jujube Based on Least Square Support Vector Machines

用可见/近红外光谱动态检测鲜枣的可溶性固形物含量。试验时样品以0.1m/s的速度运动,采集其可见/近红外漫反射光谱(350～2500nm)。用平均平滑法对120个赞皇枣样品、118个郎枣样品的光谱进行消噪处理,采用连续投影算法提取其特征波长,并建立相应的最小二乘支持向量机预测模型SPA/LS-SVM;同时将赞皇枣在500～1100nm范围的可见/短波近红外平滑光谱数据,郎枣在700～1500nm范围的平滑光谱数据用最小二乘支持向量机建立Smooth/LS-SVM预测模型,并对各自预测集样品(30个)的可溶性固形物含量进行预测和对比分析。结果表明:SPA/LS-SVM模型预测相关系数(赞皇枣0.833,郎枣0.847)与Smooth/LS-SVM模型预测相关系数(赞皇枣0.848,郎枣0.857)相差不大,且前者更精简,预测速度快,预测时间短,可以作为鲜枣可溶性固形物含量的一种动态检测方法,但模型的精度和稳定性需进一步提高。

The present research was focused on dynamic detection of soluble solids content of fresh jujube by visible and near-infrared spectroscopy. During the data gathering, fresh jujube was moving at the constant velocity of 0.1 m·s^-1 on the conveyor belt, and the visible and near-infrared diffuse reflectance spectrum （350-2 500 nm） was captured. During the data processing, moving average was employed to eliminate the spectra noise, then suc- cessive projection algorithm （SPA） was employed to extract characteristic wavelengths of 120 Fresh jujube samples named Zanhuang and 118 Fresh jujube samples named Langzao. The least square support vector machines （LS- SVM） models were established with the de-noised spectra （500-1 100 nm） and （700-1 500 nm） and characteristic wavelengths selected by using SPA. At last, the SPA/LS-SVM model and the Smooth/LS-SVM were used to predict the soluble solids content of 30 samples in the prediction set. The results showed that the correlation coefficients （0.833 and 0.847） of SPAFLS-SVM models were the same as the correlation coefficients （0.848 and 0.857） of Smooth/LS-SVM models. But the SPA/LS-SVM model was simplifier and its prediction speed was faster. So the SPA/LS-SVM model could be taken as a dynamic detection method to detect soluble solids content of fresh jujube, and the precision and stability in the model was needed to be further improved.