基于光谱信息的柑橘黄龙病无损检测及分级模型构建

柑橘黄龙病(Huanglongbing,HLB)是世界柑橘生产上最具毁灭性的病害,给果农和相关产业造成了巨大的损失.以柑橘叶片为载体,利用高光谱图像技术采集柑橘叶片表面的高光谱图像,用ENVI4.7进行图像处理,提取感兴趣区域(Region of Intest,ROI),统计感兴趣区域平均光谱数据,并进行相关植被植物的运算,最后通过PLS-DA(Partial Least Squares Discrimination Analysis)判别法进行鉴别并分类.结果表明:基于平均光谱值和植被指数的PLS-DA判别模型都能对健康、缺锌和HLB叶片进行鉴别.其中基于平均光谱值的PLS-DA模型鉴别健康柑橘叶片样品的灵敏度为100％,特异度为100％,准确度为100％;鉴别缺锌柑橘叶片样品的灵敏度为80.6％,特异度为91.7％,准确度为88.9％;鉴别HLB叶片的灵敏度为89.3％,特异度为88.3％,准确度为88.9％.基于植被指数的PLS-DA判别模型鉴别健康柑橘叶片样品的灵敏度为100％,特异度为100％,准确度为100％;鉴别缺锌柑橘叶片样品灵敏度为92.5％,特异度为89.3％,准确度为90.1％;鉴别HLB叶片的灵敏度为86.4％,特异度为95.3％,准确度为90.1％.识别正确率较高,说明利用高光谱进行柑橘黄龙病病情分类是可行的.

Citrus huanglongbing nondestructive testing and classification model construction based on spectral information

Citrus huanglongbing (HLB) is the world''s most devastating diseases of citrus production. It has caused tremendous loss to growers and related industries. Taking citrus leaves as the carrier, we collected hyper spectral images of citrus leaf surface with hyper spectral imaging technology, analyzed the image with ENVI4.7 to extract the region of interest (ROI), and counted average spectral of ROI. It operated vegetation plants related to, and finally identified and classify through PLS-DA (Partial Least Squares Discrimination Analysis) discriminant method. The results showed that three types of citrus leaves could be identified by PLSDA discriminant model based on the average spectral values and vegetation indices. Sensitivity of healthy citrus leaf samples was 100%, the specificity was 100% and accuracy was 100%; identification of zinc citrus leaf samples sensitivity was80.6%, specificity was 91.7%, accuracy was 88.9%; identification HLB leaf sensitivity was 89.3%, specificity was 88.3% and accuracy was 88.9%. For PLSDA model based on vegetation index, sensitivity of healthy citrus leaf samples was 100%, specificity was 100% and accuracy was 100%; sensitivity of zinc citrus leaf samples was 92.5%, specificity was 89.3%, accuracy was 90.1%;sensitivity of HLB leaves was 86.4%, specificity was 95.3% and accuracy was 90.1%. The correct identified rate was high. It indicates that using high-spectrum for classification of citrus huanglongbing is feasible.