Infrastructure based vehicle recognition system with multi sensor fusion

Abstract

Vehicle recognition systems are being rapidly developed. The convergence of vehicle and IT technologies conduce the vehicle recognition algorithm. There are many sensors used for vehicle detection, such as vision cameras, laser scanners, and radar and sonar sensors. There are several disadvantages when using each sensor separately, however, and we therefore propose a novel infrastructure-based vehicle recognition system using a multi-sensor fusion algorithm. We use a laser scanner and a vision camera with time and environment synchronization. Each sensor detects a vehicle, and the fusion algorithm confirms its existence and extracts its precise location information. We compare the performance of the proposed method with that of high-performance GPS data.