Matching Line Segment Scans with Mutual Compatibility Constraints

Over the years, proposals have been made to employ line segments to build 2D maps of indoor environments. One of the basic steps of these approaches is the matching between scans (or, more generally, sets) of line segments, which is usually addressed using variants of the Iterative Closest Line (ICL) paradigm. ICL is based on the idea of associating closest line segments belonging to the two scans and of reducing the distance between them. In this paper, we propose two algorithms that go beyond this approach by exploiting the mutual compatibility between associations of line segments. Experimental results show that our algorithms significantly outperform, in terms of matching accuracy, traditional algorithms based on ICL, at the cost of a slightly longer execution time.