基于高光谱成像技术的马铃薯外部缺陷检测

Detection of potato external defects based on hyperspectral imaging technology

为了实现马铃薯的准确快速分级,提出基于高光谱成像技术的马铃薯外部缺陷检测方法。通过反射高光谱成像技术采集马铃薯干腐、表面碰伤、机械损伤、绿皮、孔洞以及发芽等6类外部缺陷样本及合格样本的高光谱图像。提取合格及各类缺陷样本感兴趣区域的光谱曲线并进行光谱特性分析,采用主成分分析法确定了5个特征波段(480、676、750、800和960nm),以5个波段的主成分分析的第二主成分图像作为分类图像,识别率仅为61.52%;为了提高识别率,提出波段比算法与均匀二次差分算法相结合的方法,使缺陷识别率提高到95.65%。试验结果表明:通过高光谱成像技术可以准确有效地对常见马铃薯外部缺陷进行检测,为马铃薯在线无损检测分级提供了参考。

In order to realize accurate and fast classification of potato, a novel detection method for potato external defects was proposed based on hyperspectral imaging technology. Potatoes with dry rot, normal and other six kinds of common defects were studied. First, region of interests spectral features of various defected areas were analyzed and principal component analysis method （PCA） was used to determined five characteristic bands （480、676、750、800 and 960 nm）. Next, PCA was performed again based on characteristic bands and the second principal component was used to classify defects of potatoes, the overall classification success rate was only 61.52%. In order to improve classification success rate, band ratio algorithm and the symmetrical second difference algorithm were combined to detect external defects of potatoes. Finally, the overall classification success rate was increased to 95.65%. It is concluded that hyperspectral imaging technology can be used to effectively detect common external defects of potato.