基于支持向量机的小麦条锈病和叶锈病图像识别

Image recognition of wheat stripe rust and wheat leaf rust based on support vector machine

为了解决生产中小麦条锈病和叶锈病症状难以区分的问题,提高识别率和精度,提出了一种基于支持向量机和多特征参数的小麦条锈病和叶锈病图像分类识别方法。利用图像裁剪方法获取典型症状的子图像,采用中值滤波算法对图像进行去噪,利用K_means硬聚类算法实现病斑分割,提取病斑区域的形状、颜色和纹理特征空间的50个特征参数,设计支持向量机分类器进行分类识别。根据优选的26个特征参数,利用以径向基函数作为核函数的支持向量机对这2种小麦锈病图像进行识别。结果表明:训练样本识别率均为96.67%,测试样本识别率均为100%;与其他核函数相比,径向基核函数最适合于这2种小麦锈病的识别。所提出的基于支持向量机的方法可有效地进行小麦条锈病和叶锈病的图像识别。

It is very important to discriminate wheat stripe rust and wheat leaf rust quickly and accurately for forecast and integrated management of the diseases.In this study,a new method based on supporting vector machine（SVM） and multiple feature parameters of their images was proposed for recognition of two kinds of wheat rusts.Sub-images of visual symptoms were acquired using image cutting.The image de-noising was performed with median filtering.The diseased region was then segmented by K_means clustering algorithm.Fifty feature parameters from shape-related,color-related and texture-related features were extracted as inputs of the SVMs to identify the best classification model.The results showed that,using the SVMs with radial basis function（RBF） kernel based on the selected twenty-six features,the recognition rates of wheat stripe rust and wheat leaf rust were both 96.67% for the training sets,and 100% for the tested sets,It was thus evident that RBF kernel function was the most suitable method for image recognition of these two kinds of wheat rusts.The image recognition method based on SVM and multiple features could successfully discriminate wheat stripe rust from wheat leaf rust.