Abstract: Firstly, it established the environment model and sensor system of mowing robot, then used Otsu algorithm and fast R-cnn to segment the image and identify the obstacles. Finally,according to the mowing robot and the surrounding environment, it built a human potential field to realize the path planning of mowing robot.The experimental results show that the mowing robot can avoid the obstacles by itself and replan the preset path locally,which proves the reliability and feasibility of the system.