NCC-RANSAC: A fast plane extraction method for navigating a smart cane for the visually impaired

Abstract

This paper presents a new RANSAC based method for extracting planes from 3D range data. The generic RANSAC Plane Extranction (PE) method may over-extract a plane. It may fail in the case of a multi-step scene where the RANSAC process results in multiple inlier patches that form a slant plane straddling the steps. The CC-RANSAC algorithm overcomes the latter limitation if the inlier patches are separate. However, it fails when the inlier patches are connected. A typical scenario is a stairway with a stairwall. In this case the RANSAC plane-fitting produces inlier patches (in the tread, riser and stairwall planes) that connect together to form a plane. The proposed method, called NCC-RANSAC, performs a normal-coherence check to all data points of the inlier patches and removes those points whose normal directions are contradictory to that of the fitted plane. This procedure results in a set of separate inlier patches, each of which is then extended into a plane in its entirety by a recursive plane clustering process. The RANSAC plane-fitting and recursive plane clustering processes are repeated until no more planes are found. A probabilistic model is introduced to predict the success probability of the NCC-RANSAC method and validated with the real data of a 3D camera-SwissRanger SR4000. Experimental results demonstrate that the proposed method extracts more accurate planes with less computational time than the existing RANSAC based methods. The proposed method is intended to be used by a robotic navigational device for the visually impaired for object detection/recognition in indoor environments.