Aiming at the problem of low accuracy and poor real-time performance of obstacle detection caused by uneven density of point cloud data in the process of obstacle detection by LiDAR, an optimized density noise spatial clustering (DBSCAN) algorithm is proposed to improve the clustering effect of road obstacles. First, lane lines are detected according to the reflection intensity information of point cloud data, and regions of interest are extracted. Then, the slope thresholding algorithm is used to segment the slope ground thoroughly. Finally, an adaptive DBSCAN algorithm is proposed, which selects representative core points and adaptive clustering radius to achieve fast and accurate clustering of dense obstacles with different distances. Experimental results show that the proposed algorithm can accurately cluster the dense obstacles at different distances. Compared with the traditional algorithm, the positive detection rate of the proposed algorithm is increased by 24.07 percentages, and the average time is reduced by 1.18 s.