基于图像的水稻病害识别方法研究

Study on Recognition Method of Rice Disease Based on Image

利用图像处理技术和贝叶斯判别法对水稻3种常见病害进行识别研究。首先,利用颜色特征与病斑外轮廓分割病斑,提取病斑形态、颜色、纹理特征共63个参数;然后,应用逐步判别分析法对4个不同参数集合筛选最有效识别参数;最后,利用贝叶斯判别法进行分类识别。结果表明,逐步判别法最多可使参数减少到原来的35.2%,识别准确率最高为97.2%。该方法可以应用于其他农作物病害识别。

The recognition method of three kinds of rice diseases （blast,sheath blight,and bacterial leaf blight） was studied by image processing technique and the Bayes discrimination method. Firstly,the disease spots were segmented by color features and their outlines,and 63 parameters of shape,color and texture features were extracted. Secondly,the stepwise discriminant analysis was used to select effective recognition parameters from four parameter sets of shape,color,texture feature and all of them. Finally,the Bayes discrimination method was applied to classify and recognize the three kinds of rice diseases. The results showed that the number of texture parameters decreased to 35.2% and the highest recognition rates of four parameter sets was 97.2%. This method could also be applied to recognize diseases of other crops.