Building Extraction from LIDAR Point Cloud Data Using Marked Point Process

Abstract

This paper presents a new algorithm for building extraction from LIDAR (Light Detection and Ranging) point cloud data on the basis of a marked point process based building model. In this building model, the positions and geometries of buildings are modeled by a point process and its marks, respectively. The geometric marks for buildings include their length, width, direction, height. By Bayesian paradigm, a posterior distribution for the marked point process conditional on the LIDAR point cloud data is obtained. The Reversible Jump Markov Chain Monte Carlo (RJMCMC) based scheme is designed to simulate the posterior distribution. Finally, Maximum A Posteriori (MAP) strategy is used to obtain the optimal building detection. The proposed algorithm is tested by a set of LIDAR point cloud data. The results show its efficiency in complex residential environments.