基于高光谱成像技术的柑橘缺陷无损检测

Nondestructive Detection of Citrus Defection Using Hyper-spectra Imaging Technology

应用高光谱成像技术无损检测柑橘的缺陷。选取蒂腐、黑斑、褐腐、结痂缺陷果和正常果各30个,提取并分析了5类果皮感兴趣区域光谱曲线并结合主成分分析法确定2个最佳波长(615 nm和680 nm),然后基于特征波长作主成分分析,选取第2主成分作为分类识别图像,提出采用特征波长主成分分析法与波段比算法相结合的方法,识别率达到94%。试验结果表明,高光谱成像技术可以有效地对带有蒂腐、黑斑、褐腐、结痂缺陷的柑橘进行分类识别。

A hyperspectral imaging system was developed for detecting various common defects on citrus.Citrus with end rot,insect dot damage,rot damage,thrip scars and normal citrus were chosen.Hyperspectral images of citrus samples and principal component analysis（PCA） were used to confirm two best wavelengths（615 nm and 680 nm）.PC2 of PCA was selected to classify the images.Finally,the detection algorithm combined PCA and band ratio was developed and achieved an accuracy of 94%.The results showed the feasibility of the proposed method.