The railway system is one of the major means of transport in the transportation system, and the hidden danger to railway safety is mainly from the intrusion obstacles on the railway, such as dangerous rocks, animals, and pedestrians, and thus the technology for real-time detection of small obstacles is essential. In this paper, the 16-ray 3D lidar is used to collect data and realize point cloud imaging. The background point cloud and the point cloud collected in real time are first spatially voxelized based on the octree method, and then difference operations are performed on them. In addition, the statistical filtering and radius filtering are used to denoise the point cloud. Finally, the real-time detection of small target obstacles is realized. The research results provide some references for the data collection of a 16-ray 3D lidar scanning system.