Many of ADAS applications such as pedestrian and vehicle detection are using stereo vision. By computing the well-known disparity the range of objects ahead of the car can be determined. According to the computed range, it is possible to identify whether this object is in a collision position or not. Based on the chain code algorithm a new disparity computation method is introduced. The proposed method enhances obviously the delivered disparity values as well as the object segmentation results by integrating the segmentation into the disparity assignment step. The constitution of start pixels, which are considered as the decisive points of the disparity variation, enables object segments separation according to the assigned disparity. Furthermore it increases the reliability of the assigned disparities and thus helps the avoidance of the well-known ambiguity problem, which can be specifically encountered at image horizontal edges. Moreover the new disparity estimation method offers the on-line applicability over the traditional stereo matching approaches because fewer correlations are required.
Enhanced disparity computation for ADAS applications
Verbesserte Disparitätsberechnung für Fahrerassistenzsystem-Anwendungen
Informatik trifft Logistik, INFORMATIK, Jahrestagung der Gesellschaft für Informatik e.V., 37 ; 135-139
2007
5 Seiten, 5 Bilder, 14 Quellen
Aufsatz (Konferenz)
Englisch
Vehicle Control in ADAS Applications
Wiley | 2015
|Europäisches Patentamt | 2020
|