Fast Vanishing Point Detection based Indian Road Segmentation and Tracking

Abstract

Autonomous navigation on road is the future of road transportations. Intelligent vehicles, driver assistance systems, automatic lane departure warning system, driverless car are capturing interest of many automobile companies. With advancements in computer vision and robotics, researchers across the globe have started working in the field of vision based road navigation systems. Vision based approach is the future of intelligent vehicles. The motive of this research work is to develop a monocular vision based accurate, precise and robust road segmentation algorithm for Indian Roads. The algorithm should be universal. It should work for every road conditions i.e. highways, urban roads and rural interior roads of India. The varying near-by surroundings of rural interior roads, non-availability of lane markings and similar road and off-road regions bring additional challenges to the proposed newlineproblem. Though a good amount of work has been done in road segmentation from image, there is strong need of research work to compete against continuous developments in computer vision techniques and needs of automobile industries. This thesis discusses this assertion in detail and describes the theoretical and experimental work done to show its validity. newlineThis thesis analyses the issues connected with existing road segmentation methods and establishes a monocular vision based robust road segmentation and tracking algorithm for Indian Roads. In order to segment road region on various road types i.e. well-structured highways, semi-structured rural roads with or without lane markings and unstructured roads with similar road and off-road regions, it is proposed to make use of global road feature like vanishing point for identification of road boundaries. The proposed vanishing point based road detection algorithm overcomes the limitations of region feature based segmentation methods on varying road conditions like road types, illumination variations in day-time and presence of shadows.