Detection of Free Space/Obstacles in Front of the Ego Car Using Stereo Camera in Urban Scenes

Abstract

Transportation is one of the most important needs of a modern human being. It sustains (either directly or indirectly) the basic human needs such as food, education and his livelihood. Humans spend considerable amount of time in transit. Transportation is also not without its dangers. An estimated 1.25 million people died in the year 2013 in road traffic accidents . With the advancements in electronics, smarter and faster sensors can help reduce some of these road traffic accidents. Detection of an obstacle is an essential aspect in avoiding collisions. The aim of this thesis report is to address the challenge of road surface detection. The thesis work begins with the implementation of the v-disparity road surface estimations and proposing certain variations that offer subtle advantages. Additionally the free space estimations through 3D occupancy grid maps (OGM) have also been implemented. A novel ‘extended’ u-disparity OGM is proposed that has certain advantages to the standard OGMs. All these road surface detection algorithms are evaluated with the training datasets prepared by Karlsruhe Institute of Technology.