Vehicle localization using an AVM camera for an automated urban driving

Abstract

This paper presents a map-matching-based vehicle localization algorithm for application to automated driving on urban road. Vehicle position estimation of centimeter-level with low-priced commercial sensor setup is one of the key issue in urban automated driving. The information fusion method of localization algorithm utilizes vehicle chassis sensor and Around View Monitoring (AVM) module with four fish-eyed cameras. The proposed localization algorithm consists of three sections: a lane detection, a position correction, and a localization filter. A lane information is extracted from AVM image around the vehicle. This lane information is possible to correct vehicle position by the iterative closest point (ICP) algorithm which estimates the rigid transformation between the lane map and lanes obtained by AVM in real-time. The corrected vehicle position by this transformation is fused with the information of vehicle sensors based on an extended Kalman filter (EKF). In order to achieve higher accuracy, the covariance of the ICP algorithm is estimated by using Haralick's method. The performance of proposed localization is verified through vehicle experiments on proving ground and actual urban road.