The improved quality of transport highways and the provision of lane reflectors (cats eyes) is making it feasible to consider automatic steering systems based on computer vision technology. This paper considers the use of stereoscopic video cameras located at the front of the vehicle and pointed in the direction of lane boundary lines. In the context of viewing reflectors and cats eyes, the computer vision problem reduces to that tracking a point object in the road plane. The finite displacements in the images of the two cameras are used as an input to the computer algorithm, whose output gives an estimate of the lateral distance of the vehicle from the lane boundary. Techniques are investigated to improve the accuracy of estimating the lane boundary position which suffers from noise errors due to vibration of the camera pairs. It is shown that lowpass filtering of the position estimates is necessary, but that this approach has limitations. Conclusions and suggestions for improving the position estimates are given.
The feasibility of autonomous steering using stereo vision
Möglichkeit der autonomen Fahrzeuglenkung mit Hilfe des maschinellen Sehens
1995
5 Seiten, 4 Bilder, 3 Quellen
Conference paper
English
The feasibility of autonomous steering using stereo vision
Automotive engineering | 1995
|The Feasibility of Autonomous Steering Using Stereo Vision
British Library Conference Proceedings | 1995
|Omnidirectional stereo vision for autonomous vehicles
TIBKAT | 2015
|Omnidirectional Stereo Vision for Autonomous Vehicles
GWLB - Gottfried Wilhelm Leibniz Bibliothek | 2014
|