Review of Vehicle Detection Systems in Advanced Driver Assistant Systems

Abstract

Driverless cars and autonomous vehicles have significantly changed the face of transportation those days. Efficient use of vision system in the recent development of advanced driver assistance systems since last two decades have equipped cars and light vehicles to reduce accidents, congestion, crashes and pollution. The robust performance of the driver assistance systems absolutely depend on the flawless detection of the vehicles from the images. Developments of vigorous computer vision techniques based on various Image level features have enabled intelligent Transportation systems to solve some of the core challenges in vehicle detection. A detailed study of the vehicle detection in dynamic conditions is presented in this paper. The complexity of the vehicle detection in variable on-road driving conditions is evident from the diverse challenges illustrated in this paper. Dynamic vehicle detection mechanism has obviously attracted numerous approaches like feature based techniques and model based techniques. Different set of visual information representation as edge, shadow, light are used to detect the vehicles. Out of all low level features shape representation for vehicle detection is observed more efficient. The need of handling massive visual data for processing is addressed using novel feature representation like object proposal methods is discussed in more detail. The efficacy of ongoing research in Autonomous vehicles is validated using deep learning techniques on aerial image analysis.