园艺作物病害的多光谱组合分类  

Horticultural Plant Diseases Multispectral Classification Using Combined Classified Methods

作　　者：冯洁[1] 李宏宁 杨卫平[1] 侯德东[1] 廖宁放[2] 

机构地区：[1]云南师范大学物理与电子信息学院,云南昆明650092 [2]北京理工大学颜色科学与工程国家重点专业实验室,北京100081

出　　处：《光谱学与光谱分析》2010年第2期426-429,共4页Spectroscopy and Spectral Analysis

基　　金：国家自然科学基金项目(60968001;60768002);云南省自然科学基金项目(2009CD047)资助

摘　　要：选取设施园艺作物黄瓜的主要病害为研究对象,利用窄带滤光片型多光谱成像系统,获取患病黄瓜叶面的14个可见光通道和近红外通道、全色通道的多光谱图像。采用多光谱图像分类技术中的距离分类器、相关系数分类器和BP人工神经网络分类器,将不同病害类型对应的16个波段的反射率看作线性波谱,对210个多光谱数据样本进行识别分类,目的是探讨一个能有效识别黄瓜植株常见病害的多光谱组合分类器。实验结果表明,将人工神经网络和距离分类器有效组合,不仅分类性能明显优于单个分类器的性能,而且能够充分发挥各个分类器的特性,对园艺作物病害进行快速、准确、实时的无损检测。The research on muhispectral data disposal is getting more and more attention with the development of muhispectral technique, capturing data ability and application of muhispectral technique in agriculture practice. In the present paper, a culti- vated plant cucumber＇ familiar disease （Trichothecium roseum , Sphaerotheca f uliginea , Cladosporium cucumerinum , Corynes pora cassiicola, Pseudoperonospora cubensis） is the research objects. The cucumber leaves multispectral images of 14 visible light channels, near infrared channel and panchromatic channel were captured using narrow band muhispectral imaging system under standard observation and illumination environment, and 210 muhispectral data samples which are the 16 bands spectral re flectance of different cucumber disease were obtained. The 210 samples were classified by distance, relativity and BP neural net- work to discuss effective combination of classified methods for making a diagnosis. The result shows that the classified effective combination of distance and BP neural network classified methods has superior performance than each method, and the advantage of each method is fully used. And the flow of recognizing horticultural plant diseases using combined classified methods is presented.

关 键 词：多光谱成像技术 光谱分类 分类器 园艺作物病害 

分 类 号：O433.4[机械工程—光学工程]