Fast plane extraction method based on the point pair feature

Extracting planes from a three-dimensional (3D) point cloud is a challenging problem for many applications with 3D point clouds. In this paper, a novel fast plane extraction method based on the point pair feature (PPF) is proposed. There are two stages included in the proposed method. One is the local processing stage to sample some points in a point cloud and calculate their PPF descriptors. In this stage, the coplanar property of the PPF is used to extract initial planes from the sampling points. The other one is a global processing stage to consider all the other points in the point cloud, and assess whether they are located in the initial planes by calculating the distance from each point to the initial planes. We can extract and determine the final planes in the global processing stage. Compared with the efficient random sample consensus (RANSAC) and the 3D kernel-based Hough transform (3DKHT), the results show that for the complex scene, the extracting time of our method is less than 0.3% of the RANSAC method, and the precision rate of our method is about 9% and 17% higher than that of the RANSAC method and 3DKHT method, respectively.