基于纹理特征和支持向量机的玉米病害的识别

针对玉米病害叶片彩色纹理图像的特点，提出一种将支持向量机和色度矩分析应用于玉米病害识别的方法。首先利用色度矩提取玉米病害叶片纹理图像的特征向量，然后将支持向量机分类方法应用于病害的识别。玉米病害纹理图像识别实验结果表明：支持向量机分类方法对于病害分类训练样本较少时，具有良好的分类能力和泛化能力，适合于玉米病害的分类。不同分类核函数的相互比较分析表明，径向基核函数最适合于玉米病害的分类识别。

Recognition of Maize Disease Based on Texture Feature and Support Vector Machine

According to the features of color texture image of maize disease,a method of recognizing disease by using support vector machine (SVM) and chromaticity moments is introduced.At first,the extracting features of chromaticity moments of texture image of maize disease is done,then classification method of SVM for recognition of maize disease is discussed.Experimental results prove that SVM method fits for classification of maize disease.The comparison of different kernel functions for SVM shows that RBF kernel function is most suitable for recognition of maize disease.