Abstract: Aiming at the problems of large training convergence error and low recognition accuracy in current crop leaf disease recognition algorithms based on Machine Learning, this paper studied the intelligent detection system algorithm of plant leaf disease based on convolutional neural network CNN. Through the K-means clustering algorithm on the collected leaf images to segment the infected areas in the leaf images, and then use CNN network for feature extraction and recognition classification, to achieve the detection of crop leaf diseases. The simulation experimental data shows that, taking potato plant leaves as an example, the average recognition accuracy of the algorithm model designed in this paper is 94.7%, which is 6.15% higher than that of SVM. It is suitable for the intelligent detection system of plant leaf diseases.