基于改进颜色特征的小麦病害图像识别技术研究

Image Recognition of Wheat Diseases Based on Improved Color Feature

[目的]介绍一种根据小麦病害图像的颜色特征进行病害识别的方法。[方法]首先对小麦叶部图像进行预处理,利用小波变换进行病害部位增强和去噪;然后基于病害部位的非绿特征进行图像分割,得到只包含病害像素的图像;对病害图像颜色进行统计,得到R、G、B分量的均值,并用相对于绿色分量的均值比作为颜色特征值;最后通过分析样本图像得到每种病害的特征值范围,利用颜色特征值对未知样本进行病害识别。[结果]采用该方法对小麦叶锈病、条锈病、白粉病进行识别,平均准确率达到98%。[结论]为小麦病害的诊断与诊治提供了理论依据。

[ Objective] The aim was to introduce an image recognition method of wheat diseases based on color feature. [ Method] First,image preprocessing using wavelet transform was made for image enhancement and de-noising; then image segmentation was made based on non-green feature of disease pixels, obtaining image with only disease pixels ; mean values of R, G and B were computed to get color mean ratios as color features ; finally feature ranges were obtained by analyzing sample images, and unknown disease recognition was performed using color features. [ Resuit] The recognition accuracy of wheat diseases leaf rust, stripe rust and powdery mildew by the method reached 98%. [ Conclusion] The research provides theoretical basis for the diagnosis and treatment of wheat diseases.