Crop health status directly affects food production and foodsecurity. As one of the most important food crops in the world, maize disease prevention and control is particularly important. Traditional manual detection is time-consuming and laborious, and its accuracy is limited. In recent years, the application of image recognition technology based on deep learning in the agricultural field provides a new way for maize disease recognition. The purpose of this paper is to explore a convolution based Neural Network (Convolutional Neural Network, CNN) of maize disease classification system, through the collection, with a large number of corn leaf images, training high performance recognition model, to achieve accurate identification and classification of several kinds of common maize disease. The system can not only greatly improve the speed of disease detection, but also reduce the misdiagnosis rate and provide real-time and effective technical support for agricultural production.