塔里木大学学报

以库尔勒香梨的含糖量作为研究和检测指标，使用便携式近红外光谱仪采集香梨样本光谱数据，采用一阶差分、二阶差分、标准正态变量变换(SNV)、多元散射校正(MSC)等预处理方法对原始光谱进行预处理分析，研究香梨糖分的近红外光谱响应，并使用相关系数法提取12个特征波长变量，根据库尔勒香梨标准，以糖度特征光谱数据作为参数，利用最近邻域法(KNN)、支持向量机(SVM)、随机森林(RF)方法建立库尔勒香梨等级判别模型。结果表明，KNN模型的分类结果优于其它两种预测模型，可用于构建基于近红外光谱的库尔勒香梨等级评判模型。MSC+KNN处理方法可用于构建库尔勒香梨等级评判模型，为进一步研究库尔勒香梨等级评判的便携式检测装置提供理论参考。

Precise classification and grading by Korla pear's sugar index plays an important role in the process of quality inspection. This paper adopted infrared spectrometer to achieving the classification of Korla pear. The spectral data of pear samples were collected by portable near infrared spectrometer, and the grade discrimination model of Korla pear was established combined with chemometrics. The original spectra was pre-processed and analyzed by first-order difference, second-order difference, standard normal variable transformation(SNV) and multivariate scattering correction(MSC). The near-infrared spectral response of pear sugar was studied, and 12 characteristic wavelength variables were extracted by correlation coefficient method. According to Korla pear standard, the sugar feature spectral data was used as parameters. The classification model of Korla pear was established by k-nearest neighbors(KNN), support vector machine(SVM) and random forest(RF). Experimental results show that the classification result of KNN model is better than the other two prediction models based on sugar degree. MSC+KNN processing method can be used to construct the grade evaluation model of Korla pear.