摘要 |
The first step in processing three dimensional range data for a vision system of a robot is feature extraction such as planar segmentation. When the sensors used to obtain the range data are mounted on a platform which moves during data acquisition, inter scanline connectivity cannot be relied upon. Therefore, line segments are detected and merged in each scanline and then line segments of different scanlines are merged to form planes. Both merging steps use a least squares threshold. To speed processing, the viewing area is broken into cells and comparisons are made only between cells adjacent to a cell of interest. The resulting planes undergo conventional hypothesis generation and verification.
|