Multidimensional particle swarm optimization-based unsupervised planar segmentation algorithm of unorganized point clouds

This paper presents an unsupervised planar segmentation algorithm of unorganized point clouds based on multidimensional (MD) particle swarm optimization (PSO). A robust objective function of the unsupervised planar segmentation is established according to clustering distances of PSO clustering algorithm and inliers of random sample consensus (RANSAC) method. After that, MD PSO algorithm is adopted to optimize the objective function, where the optimal number and positions of the segmented planar patches are sought simultaneously. In order not to get trapped in local optima, a modification strategy of the global best (GB) position of swarm in each dimension is added to the MD PSO algorithm. Thus the unsupervised planar segmentation of point clouds is realized. Experimental results demonstrate the high planar segmentation accuracy of the proposed algorithm.

► An unsupervised planar segmentation algorithm of unorganized point clouds based on multidimensional PSO is proposed.
► A robust objective function of unsupervised planar segmentation is established.
► The number and positions of planar patches are optimized simultaneously.
► The global best position of swarm is modified to avoid trapping in local optima.
► Our algorithm yields higher accuracy compared with other approaches.