Finding Multiple Lines from 2D Range Data: An Odds Ratio Approach

C. Lara and L. Romero (Mexico)

Keywords

Robot Sensing, Line Maps, SLAM, Line Features.

Abstract

Features are very important for high level robot tasks such as localization, mapping, place recognition, or navigation. Line features are useful for man-made environments such as rooms and corridors. There are several solutions for ex tracting multiple lines from range scans. Bottom–up tech niques obtain local clusters and then similar clusters are merged; usually, a distance between each pair of current clusters is computed and the closest pair is merged iter atively until a stopping criterion is met. This paper pro poses an odds ratio test for deciding when two linear clus ters must be merged or not; this test is a good alternative to the classical approaches based on a distance criterion. Experimental results show that the proposed method find reliable lines.

Important Links:



Go Back