An algorithm for the automatic segmentation of dense circular pipeline point cloud data is proposed. The cloud data is divided into several sub-blocks based on the octree structure, among which the spatial neighborhood relationship is established. The random sampling consensus algorithm based on the normal vector constraints is used to remove the large area plane within each sub-block and simultaneously, the Euclidean distance clustering and the region growing segmentation algorithm based on the smoothness constraints are used to refine the data again. The experimental results show that a 4 thread parallel computation only takes 9 s and the precision is larger than 90% when the proposed automatic segmentation algorithm is used to process the data with a size of 6 m×12 m×16 m in the point cloud space. Thus the proposed algorithm can be used for the quick and accurate segmentation of pipeline point cloud data and has a high application value.