Generic obstacle detection on roads by dynamic programming for remapped stereo images to an overhead view

This paper proposes an approach to extract generic obstacles on roads using remapped stereo images to an overhead view. The proposed approach uses the fact that the information of road surface on the remapped image is distorted by an obstacle, and formulates the generic obstacle detection (GOD) problem as a dynamic programming (DP), which contributes to search for corresponding peaks on polar histograms constructed by vertical edge components of the remapped stereo images. The corresponding peaks determine the estimates of obstacles' positions. The approach has features that it is not largely affected by intensity difference between a pair of stereo images and does not depend on the typical stereo matching, and identifies the obstacle's position quite well.