PROJECT TITLE :

On Enhancing Lane Estimation Using Contextual Cues

ABSTRACT:

Vision-based mostly lane detection is a crucial part of modern automotive active safety systems. Although a range of strong and accurate lane estimation (LE) algorithms have been proposed, computationally efficient systems which will be realized on embedded platforms are less explored and addressed. This paper presents a framework that comes with contextual cues for LE to more enhance the performance in terms of each computational efficiency and accuracy. The proposed context-aware LE framework considers the state of the ego vehicle, its surroundings, and therefore the system-level needs to adapt and scale the LE method ensuing in substantial computational savings. This is often accomplished by synergistically fusing data from multiple sensors along with the visual information to outline the context around the ego vehicle. The context is then incorporated as an input to the LE process to scale it depending on the contextual necessities. A detailed evaluation of the proposed framework on real-world driving conditions shows that the dynamic and static configuration of the lane detection process leads to computation savings as high as ninety%, while not compromising on the accuracy of LE.