可见/近红外光谱技术无损识别苹果品种的研究

为了能快速无损鉴别苹果的品种,基于高光谱成像技术结合模式识别算法,分别建立了判别苹果品种K最近邻(KNN)识别模型与偏最小二乘判别分析(PLS-DA)识别模型。综合比较了不同光谱预处理方法(二阶微分(SD)、标准正态变换(SNV)和多元散射校正(MSC))对各模型识别效果的影响,并利用主成分分析方法对预处理后的光谱数据进行降维,以提取能反映苹果品种的特征光谱。结果表明:采用主成分分析法选择了累计贡献率超过99.9%的前10个主成分作为样本集特征光谱数据,很好地实现了光谱数据的降维;MSC预处理方法对光谱反射率预处理的效果最好;2种判别模型均能基本满足实际要求,且MSC+KNN识别模型的识别性能最优,对预测集样本的正确识别率高达100%。

Nondestructive Identification of Apple Varieties by VIS/NIR Spectroscopy

In order to achieve rapid nondestructive identification of apple varieties, the recognition models of apple varieties were established based on hyperspectral imaging technology combined with K Nearest Neighbor (KNN) and Partial Least-Square Discriminant Analysis (PLS-DA), respectively. Then the effectiveness of the discriminant model using Second Derivation (SD), Standard Normal Variation (SNV) and Multi-Scatter Calibration (MSC) was compared and evaluated. Finally, the characteristic spectrums of apples were extracted by the Principal Component Analysis(PCA). The results showed that, the first 10 principal components with cumulative contribution rate of 99.9% were selected by the principal component analysis as the characteristic spectral data in the sample set, and the dimensionality reduction of the spectral data was well realized. The preprocessing effect of MSC on spectral reflectivity was the best. Both models could basically meet the practical requirements, and MSC+KNN model had the optimal recognition performance with an accuracy recognition rate of 100%.