Road Boundary Extraction from LiDAR Point Clouds Using Dynamic Programming Algorithm

In order to solve the problem that it is difficult and inaccurate to extract road boundaries from laser point clouds, a road boundary extraction method based on a dynamic programming algorithm is proposed. First, the point clouds are projected, and an obstacle grid map is obtained using the dual height difference algorithm to retain the ground features near the curb. Considering the distance from the boundary points to the driving trajectory, the dynamic programming algorithm is used to extract the road boundary. Starting from the driving trajectory, the algorithm searches for the white obstacle points in the grid maps on both sides, and the point with the minimum total cost is considered the boundary point. This study used the IQmulus & TerraMobilita datasets for experiments. The results showed that the accuracy and recall rates of the extraction results from both datasets were above 98.7%, and the F1 scores were both above 98.4%.