Environment perception and analysis are essential parts of automatic ground vehicles (UGVs) to implement smart driving-decision making. To sense dynamic environment information, light detection and ranging (LiDAR) are equipped on UGVs to collect high-precision 3-dimension point cloud. Because of the unstructured and inhomogeneous characteristics of LiDAR point cloud, the fast and accurate analysis of point cloud is hard to achieve under UGVs’ driving. Most UGVs’ autonomous applications, such as object extraction, terrain perception, and traversable path recognition, face technology bottleneck in both process speed and analysis precision. In these applications, object segmentation result is a fundamental information support, which influence the subsequent processes to a large extent in both accuracy and efficiency performance. This paper proposed a novel object segmenting algorithm named 3D connected component label (3D-CCL) to divide full point cloud into subdivision point clouds of individual local object space. The object segmenting result provide a series of basic point clouds of different obstacle models, which benefits for the environment perception and decision making for UGV.