Novel approach to lane and path detection in unmanned ground vehicles

Abstract

The dire need for reduction in the road deaths has led to a massive development in the field of autonomous vehicles. Autonomous vehicles are a niche market due to the cost involved. To reduce the costs and facilitate the use of autonomous vehicles to the common man, reduction in number of sensors and improvements in the software is desired. This has led to the penetration of the computer vision into the field of autonomous vehicles. In this paper, we present a novel algorithm for lane detection called Vertical Sub-band Spectrum Extraction (VSSE) for urban autonomous vehicles. Also, we propose a novel path detection algorithm called Adaptive Color band Thresholding (ACT) for pastoral areas. Both the algorithms are intended for the purpose of using minimum hardware for real time operation. Experimental results, obtained by applying the proposed algorithms on varying conditions of illumination, markings and other background variations for still images as well as real time video feed, show that the proposed system is in par with more complex systems. A significant detection rate under varying conditions with a relatively simple hardware is observed.