A global line extraction algorithm for indoor robot mapping based on noise eliminating via similar triangles rule

Abstract

Robot mapping from 2D laser rangefinder data is critical to indoor autonomous mobile localization and navigation. On the basis of Split algorithm, this paper presents a similar triangles rule, which is used to quantify the random characteristics of the space position of noise points. The direction of the real line is acquired by using the statistical method to analyze the quantified data, and the noise points can be eliminated based on the actual line direction as much as possible. Finally, we split the points after denoising again and then fit lines by least square method in each two adjacent split points. Experiments show that, compared with the conventional Split-and-Merge algorithm, the proposed technique performs better on both accuracy and correctness. The process of line segments merging is avoided, and the precision and robustness of environment modeling are enhanced in the proposed algorithm.