Interpretation And Integration Of Height Maps From A Range Image Sequence

This paper presents a method for interpreting and integrating of height maps from a range image sequence taken by a mobile robot in a road scene. The height map represents the height information, which is transformed from a range image, in the vehicleccntcretl Cartcsiari coordiiintc systciii. I;irst, the height mxp is segmented into unexplored, occluded, traversable and obstacle regions from the height information, then, obstacle regions are classified into artificial objects or natural objects according to their geometrical properties such as slope and curvature. Next, the system matches height maps produced at different observation stations in order to combine them, finding correspondence of regions between adjacent height maps. A special care needs to be taken for moving objects because they have different motion parameters from those of the stationary environment. Large errors in matching two obstacle regions corresponding to each other occur at the boundary of observable area and at the area with bad range data where the height information is inconsistent with the ranging geometry. Geometrical reasoning for them can reduce the large matching errors, and as aresult, the system can derive the correct motion parameters. Finally, the system integrates the height maps into a local map using the obtained motion parameters and updates region labels of the loc:il map for matching with the following height maps. We show the results applied 10 l;indscnpe inodcls using ALV simulator of the University of Maryland.