Planar Surfaces Recognition in 3D Point Cloud Using a Real-Coded Multistage Genetic Algorithm

Most frequent surface shapes of man-made constructions are planar surfaces. Discovering those surfaces is a big step toward extracting as-built/-is construction information from 3D point cloud. In this paper, a real-coded genetic algorithm (GA) formulation for planar surfaces recognition in 3D point clouds is presented. The algorithm developed based on a multistage approach; thereby, it finds one planar surface (part of solution) at each stage. In addition, the logarithmically proportional objective function that is used in this approach can adapt itself to scale and spatial density of the point cloud. We tested the proposed application on a synthetic point cloud containing several planar surfaces with different shapes, positions, and with a wide variety of sizes. The results obtained showed that the proposed method is capable to find all plane’s configurations of flat surfaces with a minor distance to the actual configurations.