Abstract: In order to improve the accuracy of machine learning algorithms in recognizing sliding-like discharge patterns, this paper proposes a method based on convolutional neural network (CNN) and visual images which recognizes corona discharge, diffuse discharge, and gliding-like discharge patterns. The image library is constructed by selecting sliding-like discharge images under different conditions such as a gas flow rate of 0~16 L/min, an electrode gap of 2~10 mm, and a pulse frequency of 0.5~3 kHz. Then CNN is built, and the hyperparameters which affect the recognition performance of CNN are optimized, including the number of the network layer, the number of the neuron in the Full Connection (FC) layer, the size of the convolution kernel, and the type of the activation function. Finally, the recognition performance of CNN and Decision Tree (DT) algorithm and Random Forest algorithm (RF) algorithm is compared. The results show that the accuracy of CNN recognition is 100%, which is higher than that of the traditional machine learning method. In addition, the discharge mode and the condition parameters are given. Through the clustering analysis algorithm based on Back Propagation Neural Networks (BPNN), the diffuse discharge and sliding-like discharge are identified, and the accuracy is 100%.