In order to improve the navigation obstacle avoidance ability to the substation inspection robot, a deep convolution neural network based on deep learning technology is used to robot obstacle avoidance. This method combines image classification and semantic segmentation to assist the robot in navigation and obstacle avoidance. The classification branch obtains global image information to ensure the correct direction of the robot. The semantic segmentation branch guides the robot to avoid obstacles accurately based on the local information of the image and the target category in front of the robot. Experimental results show that the method proposed in this paper can classify and segment images effectively. At the same time, in the actual substation environment, it can also provide effective obstacle avoidance information for inspection robots and realize real-time autonomous obstacle avoidance.