【摘要】 为了实现苹果表面损伤和早期腐烂的快速有效检测,利用高光谱成像和主成分分析变换挑选了检测损伤和早期腐烂的特征波段,并开发了基于特征波段的检测算法。首先,利用高光谱成像系统采集苹果可见-近红外(400～1000 nm)波段的图像;其次,基于全波段图像的主成分分析变换,挑选出6个特征波段(560、640、675、720、810和970 nm);再次,对特征波段进行主成分分析变换,挑选出第3主成分(PC3)作为检测苹果表面损伤和早期腐烂的图像。利用该算法对正常果和带有损伤和早期腐烂的苹果样本共计120个进行检测,检测正确率达到95.8%。试验结果表明:基于特征波段的主成分分析变换可以有效检测苹果表面的损伤和早期腐烂。更多还原

【Abstract】 To detect bruises and early decay in apples fast and efficiently, a hyperspectral imaging system and principle component analysis(PCA) transform were used to select the effective wavelengths, and a classification algorithm was developed based on the effective wavelengths. Firstly, the hyperspectral images in the visible and near-infrared(400-1000 nm) ranges were acquired; Secondly, six wavebands(560, 640, 675, 720, 810 and 970 nm) are selected as the effective wavebands based on principle component analysis transform for the hyperspectral images; Thirdly, the third component(PC3) scores image was selected as the suitable image for detecting the bruises and early decay in apples based on principle component analysis transform for the effective waveband images. For the investigated 120 samples, the results with a 95.8% overall detection rate indicated that the PCA transform based on effective wavebands is effective.更多还原