模式识别技术及数据挖掘方法已成为化学计量学的研究热点。 近红外(NIR)光谱分析以其快速、 简便、 非破坏性等优势广泛应用于光谱信号的处理和分析模型的建立。 基于五种不同的模式识别方法: 局部线性嵌入(LLE), 小波变换(WT), 主成分分析(PCA), 偏最小二乘(PLS)和支持向量机(SVM), 利用NIR技术建立了玉米种子的模式识别系统, 并将其应用于108玉米杂交种和母本178种子的近红外光谱样品。 首先利用LLE, WT, PCA, PLS进行消噪或降维, 然后运用SVM进行分类识别, 而一模支持向量机(1-norm SVM)算法直接进行分类识别。 三个不同NIR 光谱范围的数值实验显示: PCA+SVM, LLE+SVM和PLS+SVM 识别效果甚佳, 而WT+SVM和1-norm SVM方法也有较高的分类精度。 实验结果表明了本文提出方法的可行性和有效性, 为利用近红外光谱和模式识别技术进行种子识别研究提供了理论依据和实用方法。

Pattern recognition technology and data mining methods have become a hot topic in chemometrics. Near infrared (NIR) spectroscopic analysis has been widely used in spectrum signal processing and modeling since it has advantages of quickness, simplicity and nondestructiveness. Based on five different methods of pattern recognition, namely the locally linear embedding (LLE), wavelet transform (WT), principal component analysis (PCA), partial least squares (PLS) and support vector machine (SVM), the pattern recognition system for corn seeds was proposed using NIR technology, and applied to classification of 108 hybrid samples and 178 female samples for corn seeds. Firstly, we get rid of noise or reduce the dimension using LLE, WT, PCA, PLS, and then use SVM to identify two-class samples. In the meantime, 1-norm SVM is the method of direct classification and identification. Experimental results of three different spectral regions show that the performances of three methods: PCA+SVM, LLE+SVM, PLS+SVM are superior to WT+SVM and 1-norm SVM methods, and obtain a high classification accuracy, which indicates the feasibility and effectiveness of the proposed methods. Moreover, this investigation provides the theoretical support and practical method for recognition of corn seeds utilizing near infrared spectral data.