本文利用高光谱图像技术对干制后的哈密大枣可溶性固形物含量（SSC）进行预测研究。使用多种预处理方法对原始光谱进行处理，并对原始光谱和预处理后的光谱分别建立PLS模型，对比分析得出均值中心化（MC）处理效果最佳。对MC处理后的光谱经联合区间偏最小二乘算法（si-PLS）筛选后，再结合遗传算法(GA)和竞争性自适应重加权算法（CARS）提取哈密大枣SSC的特征波长，将提取的波长变量建立哈密大枣SSC的PLS预测模型。结果显示：利用MC-CARS-GA-si-PLS方法提取的16个关键波长变量（仅占全光谱变量的2%）所建立的PLS模型性能优于全光谱PLS模型。该模型的预测集相关系数(Rp)、预测均方根误差(RMSEP)和预测(RPD)分别为0.93、0.48和2.721。该方法提取的波长变量所建立的预测模型，不仅使模型简化，而且增强了模型的预测能力，为高光谱图像技术对水果及其干制品的定量分析研究提供了参考。

Prediction of soluble solids content (SSC) of dried Hami big jujubes was studied by means of the hyperspectral imaging technology in this work. Many spectral preprocessing methods were used to preprocess the raw spectra, and the partial least squares (PLS) models were established based on the raw spectra and preprocessed ones. The comparison and analysis showed that the best preprocessing result was achieved by the mean centering (MC) algorithm. MC-pretreated spectra were screened by the synergy interval partial least squares (si-PLS) method, and the characteristic wavelengths of SSC of Hami big jujubes were selected using a combination of a genetic algorithm (GA) and competitive adaptive reweighted sampling algorithm (CARS). The selected wavelength variables were used to build a PLS predictive model for SSC of Hami big jujubes. The results indicated that 16 characteristic wavelengths were selected and accounted for only 2% of full spectral variables in the MC-CARS-GA-si-PLS model. The performance of the newly built PLS model was better than that of the PLS model based on the full spectrum. The correlation coefficient of the prediction set (Rp), the root mean square error of prediction (RMSEP), and the relative prediction deviation (RPD) of this model were 0.93, 0.48, and 2.721, respectively. The results show that this PLS predictive model based on the wavelength selection not only is simpler but also enhances the predictive ability of such models, and can serve as a reference for quantitative analysis and study of fruits and dried fruits by the hyperspectral imaging technology.

国家科技支撑项目（2015BAD19B03）