Estimation of Ego-Vehicle’s Position Based on Image Registration

Abstract

This paper describes an estimation method of ego-vehicle’s position using a 2D map. The general approach for localization is to match the white lines between the map and the real world. However, such approaches suffer from changing the environmental conditions and painting clearance. If considerable changes occur, the false-detection rate increases due to noises and miss-detected lines. In order to solve this issue, a localization method based on the holistic road area detection is proposed. First, the road area is extracted from a predefined 2D boundary map. Next, the real world road area is detected using LiDAR and converted into the binary image plane. Finally, an image registration technique is applied to calculate the overall matching score of the road area between the map and LiDAR images. The proposed method has provided an accurate estimation against environmental changes with low-cost calculation based on the simulation results. In addition, the validation of the proposed method in the real world has performed less than 0.2 m for the estimation error.