基于近红外光谱技术的空苞山核桃快速识别

空苞山核桃是指果实没有种仁或者发育受阻的山核桃，严重影响山核桃产品品质。为实现空苞山核桃的快速无损识别，利用2种近红外检测装置在200～1 160 nm波长范围采集带壳山核桃样本的光谱，采用8种预处理方法进行光谱预处理，利用竞争自适应重加权采样（CARS）方法筛选空苞山核桃的特征波长变量，最后应用线性判别分析（LDA）、二次判别分析（QDA）和马氏距离判别分析（MDA）建立空苞和正常山核桃的分类模型。结果表明，使用检测装置1所建立的空苞山核桃分类模型性能优于检测装置2的分类模型，经多元散射校正（MSC）预处理后建立的分类模型的识别结果最好，LDA、QDA及MDA模型的特异性、敏感性和正确率均为1，优于其它预处理方法建立的分类模型。经CARS变量筛选后，建模所用的光谱变量数目大大减少，有效简化了分类模型，而模型性能仍与全波长模型性能持平。本文为空苞山核桃的快速、无损识别提供了一种可行的方法。

Rapid Identification of Empty Bud Carya cathayensis Based on Near-infrared Spectroscopy

Empty bud Carya cathayensis refer to Carya cathayensis with no seed kernels or stunted development， which will seriously affect the product quality of Carya cathayensis. To achieve rapid and non-destructive identification of empty bud Carya cathayensises， two NIR detection devices were used to collect spectra of shelled Carya cathayensises samples in the wavelength range of 200-1 160 nm. Eight different preprocessing methods wers used for spectral preprocessing， and competitive adaptive reweighted sampling (CARS) was used to select the sensitive wave length variables of empty bud Carya cathayensises. Finally， linear discriminant analysis (LDA)， quadratic discriminant analysis (QDA)， and mahalanobis discriminant analysis (MDA) were applied to establish the classification models of empty bud and normal Carya cathayensises. The results showed that the performance of the classification models of empty bud Carya cathayensis established using detection device 1 were better than that of the detection device 2. The classification models established after multivariate scattering correction (MSC) pretreatment showed the best recognition results. The specificity， sensitivity and correctness of LDA， QDA， and MDA models were all 1， which were better than the classification models established by other preprocessing methods. After CARS variable screening， the number of spectral variables used for modeling was significantly reduced， effectively simplifying the classification model. However， the model performance was still on par with that of the full-band model. This paper provides a feasible method for the rapid nondestructive identification of empty bud Carya cathayensis by near-infrared spectroscopy combined with chemometric methods.