Abstract

In 3D reconstruction and robotics, point cloud registration is a critical component of many tasks including the estimation of sensor motion. The Iterated Closest Point (ICP) algorithm and its variants were initially used to solve such problems. However ICP based methods often fail to converge to the correct solution in the presence of detection as well as spatial errors. This is because ICP methods typically use an L2 metric as part of their optimization criteria, which is unable to penalize cardinality errors. This article therefore presents a registration technique based on the multi-object Cardinalized Optimal Linear Assignment (COLA) metric, which penalizes both detection and spatial errors. This allows robust scan registration to take place in the presence of both unknown inter-scan translation and orientation as well as point cloud detection errors. The resulting Particle Swarm Optimization (PSO)-COLA registration algorithm is shown to outperform state of the art local and global point cloud registration algorithms in the presence of data outliers and spatial uncertainty.