Abstract: A naive Bayesian classifier method is proposed to classify maize leaf disease according to five kinds of actual maize disease images, which are segmented and extracted feather from at first. The result shows that the precision of maize disease identifying is higher than 83%. Bayesian classifier is excellent at simple network structure and extending easily. It is effective for classifying maize disease and it can use for reference for the image recognition research on other crops.

摘要： 本研究根据锈病、弯孢菌病、灰斑病、小斑病及褐斑病等五种玉米病斑图像的实际情况，在图像分割和特征提取的基础上，利用朴素贝叶斯分类器的统计学习方法，实现玉米叶部病斑的分类识别。研究结果表明，对五种玉米叶部病害的诊断精度在83%以上。贝叶斯分类器具有网络结构简单、易于扩展等特点，对玉米叶部病害的分类识别效果较好，也为其它作物病害图像识别的研究提供了借鉴。