In order to effectively obtain the road information about the road ahead and apply it to the low-cost solid-state Lidar, a road edge and obstacle detection method is proposed in this paper. Firstly, ground filtering processes the original point cloud data to extract the ground and non-ground point cloud data. Based on the features of high mutation in the pastry cloud data, this paper proposes to extract the path feature points by dynamic sliding window, then uses random sampling consensus algorithm(RANSAC) to fit the inner obstacle point cloud as the region of interest(ROI), treats the obstacle point cloud data with H straight pass filter in the z-axis direction, and finally uses the European clustering algorithm. This paper verifies the feasibility of the method by conducting practical data acquisition and processing experiments on campus.