Detection of Planar Surfaces Based on RANSAC and LAD Plane Fitting

摘要

In this paper, a RANSAC-based approach for the detection of dominant planar surfaces in range images obtained by stereo vision is proposed. Plane hypotheses are generated by Least Absolute Deviation (LAD) plane fitting to points taken from the neighborhood of a randomly sampled point. This way, the probability that the points used to instantiate a plane lie on the same surface is increased. Furthermore, LAD plane fitting is used instead of commonly used Least-Squares fitting since it is much more robust to outliers. The proposed technique is evaluated using both synthetic data and real data obtained by a stereo camera system mounted on a mobile robot.