Abstract: This paper selects four common diseases of wheat leaf images, using image processing techniques to identify different types of disease. Firstly, the RGB color space is converted to HSV color space, the relevant color characteristics（hue and saturation）are extracted, and then geometry features（perimeter area, squareness, roundness, eccentricity, etc.） are extracted. To obtain the eigenvalues of each disease range, the sample images are analyzed, and then the eigenvalues of the unknown samples are used to identify different kinds of wheat diseases. This research takes powdery mildew and rust（leaf rust, stripe rust and stem rust） as research objects. Based on color characteristics, the powdery mildew and rust are identified, according to the shape characteristics, leaf rust, stripe rust and stem rust are identified. The proposed method is simple and convenient with an identification rate of more than 96%. The experimental results show that the chosen color and shape features of these four common diseases are valid and feasible for wheat diseases identification.

Key words: wheat disease, feature extraction, image recognition

摘要： 选取小麦叶部常见病害图像，利用图像处理技术进行病害种类的识别。将图像由RGB彩色空间转换到HSV颜色空间，提取相关的颜色特征（色相和饱和度），接着提取几何形状特征（周长、面积、矩形度、似圆度、偏心率等），通过分析样本图像得到每种病害的特征值范围，利用特征值对未知样本进行病害识别。系统以白粉病和锈病（叶锈病、条锈病和秆锈病）为研究对象，根据颜色特征对白粉病和锈病加以识别，然后根据几何形状特征对叶锈病、条锈病和秆锈病进行识别，操作简单方便，识别准确率达96%以上。实验结果表明，选取的颜色特征和几何形状特征对4种小麦叶部常见病害的识别是有效且可行的。

关键词: 小麦病害, 特征提取, 图像识别