[Objective]The study aimed to research the automatic recognition of Rhkocotonia Solani by support vector machine (SVM) so as to make up the defect of artificial recognition and increase the accuracy and efficiency of recognition. [ Method] With R. Solani as the studied object, firstly,the method based on the vector median filtering was used to pre-treat the image of R. Solani,then the fuzzy c-mean clustering method was used to make for the gray image segmentation in the image segmentation stage and the feature parameters which represented the lesion were extracted from three aspects such as color,texture and shape,finally the SVM recognition method was used to identify Ft. Solani and was compared with the recognition method based on BP neural network. [ Result]Tlie SVM recognition method showed the recognition rate of 95.00% .which is better than that of BP neural network (91.88% ). [ Conclusion ] The recognition of R. Solani based on SVM could not only make up the defect of artificial recognition,but also increase the accuracy and efficiency of recognition,which showed the broad application prospects.%[目的]研究支持向量机对纹枯病病害进行自动识别,弥补人工识别的缺陷和不足,提高识别的准确性和效率.[方法]以水稻纹枯病为研究对象,使用基于矢量中值滤波的方法对水稻纹枯病图像进行预处理.利用模糊C均值聚类法,在图像分割阶段进行灰度图像分割；分别从颜色、纹理和形状3个方面提取代表病斑的特征参数.最后用支持向量机识别方法进行水稻纹枯病识别,并与基于BP神经网络的识别方法进行对比.[结果]识别率达到95.00％,要优于BP神经网络的91.88％.[结论]基于支持向量机的水稻纹枯病识别弥补了人工识别的缺陷,也提高了准确性和效率,有广阔的应用前景.
展开▼
