During lidar detection of obstacles, owing to the characteristics of near dense and far sparse point clouds, the movement with variable speeds of vehicles results in point cloud drifting in the object segmentation. Moreover, objects close to each other are difficult to be segmented, resulting in omissions or incorrect detections. To address these problems, this study proposes an improved Euclidean clustering algorithm based on the point cloud shot-line angle constraint to make obstacle detection more rapid and accurate. The proposed algorithm effectively solves the problem of low success rate in detecting obstacles owing to the uneven point cloud density. Simultaneously, experiments are performed on the proposed algorithm. The experimental results show that the proposed algorithm can quickly and accurately segment and cluster obstacles within a certain range compared with the traditional Euclidean clustering algorithm.