Application of Monocular Camera for Vehicular Localization and Forward Collision Warning

Abstract

Due to the vigorous development of intelligent vehicles the accurate vehicle state is an indispensable element for vehicle dynamic stability system collision avoidance systems lane departure warning system and even the automatic navigation system Generally the global positioning system (Global Positioning System GPS) can be used to obtain the vehicle position direction and speed However the positioning accuracy becomes not credible enough when GPS is disturbed the accuracy of vehicle status can be increased by an inertial navigation system (Inertial Navigation System INS) and image sensors In this thesis the vehicle kinematic model and the extended Kalman filter algorithm are adopted to develop a multi-sensors fusion algorithm for precise positioning With the popularity of driving recorder almost every vehicle is equipped with a driving recorder The aim of this research is to estimate vehicle state by fusing the measurements gathered from inertial measurement unit (IMU) GNSS receiver and image sensors In addition the vision-based collision warning system can warn the driver according to the forward vehicle width while the inter-vehicle distance gets too close Image processing technique is implemented in LabVIEW user machine interface (UMI) which can be observed in real time The proposed algorithm is verified by using real gathered data in Matlab platform The experimental results demonstrate not only that the accuracy of the vehicle state is improved by the proposed positioning algorithm in GPS challenging environments but also the proposed forward collision warning system is feasible

Date of Award	2014 Jul 30
Original language	English
Supervisor	Jyh-Chin Juang (Supervisor)