A fast and efficient algorithm to build an environment map ispresented for the stereo vision of indoor mobile robots, by using imagereduction and parallel processing. The real images of corridors arecomplex images and may have weak contrast under natural lightingcondition, so large reduction may spoil the accuracy of an environmentmap. To keep the accuracy of the map, the pyramid in this paper has onlytwo levels. To reduce the number of computations minimally, itsreduction scale can be any size including 1/2n. Also, theparallel processing is used for fast computation. A data structure ofline support region and the fast merging algorithm are proposed. Theproposed algorithm is compared with an algorithm that uses only originalimages. The experimental results show the proposed algorithm savesprocessing time, which takes only 7% of the conventional algorithm, andpreserves the accuracy of an environment map
展开▼
