摘要
在研究多光谱成像技术特点的基础上,提取光谱有效特征信息,优化光谱编码条件,对4个等级的黄瓜霜霉病害叶面光谱进行融合编码分类识别,可以有效地降低光谱的相似性,增强可分性,实现了4个病害等级的较好分类。在编码识别的基础上,采用一种权衡性较强的评价方法对分类结果进行评估,这种评价方法实现了把病害分类由4个离散的等级扩展到从0到1的连续量化的评定,并且可以精确到对单个像素点的等级评估。
Based on the study of Multi-spectral imaging technology,we extracted spectral effective feature information and optimize coding conditions.In this paper,a fusion coding for spectrum method was presented.Classification and recognition of four levels of cucumber downy mildew was achieved by fusion coding for the spectrum of foliar disease.The result of classification showed that the spectral similarity was reduced effectively and the separability was enhanced by fusion coding.Then,an evaluation method was adopted to evaluate the result of classification in this paper.The evaluation method expanded disease rating levels from four discrete classificationlevel to continuous and quantifiable level ranging in between 0 and 1.The quantification accuracy reached to the pixels level.
出处
《云南师范大学学报（自然科学版）》
2011年第6期63-69,共7页
Journal of Yunnan Normal University:Natural Sciences Edition
基金
国家自然科学基金资助项目(60968001
60768002)
云南省自然科学基金资助项目(2009CD047)
云南省大学生创新实验(CX07)
关键词
多光谱成像技术
光谱特征
光谱编码
加权归类
病害评估
Multispectral imaging technique
Characteristic spectrum
Spectral encoding
Weighted classifier
Disease level evaluation
