Ground Texture Matching based Global Localization for Intelligent Vehicles in Urban Environment

Localization is a critical problem in the research of intelligent vehicles. Traditional vision based methods suffer from reliable problems in outdoor environment, especially in complicated urban areas. This paper proposes a novel approach to localize the position of a vehicle with respect to a global map, based on the texture of the ground where the vehicle moves. Images of the ground texture are obtained from a camera viewing downward fixed on the bottom of the vehicle. Thus, the vehicle actuates like a big optical mouse. Global localization is achieved by map matching using ICP (iterative closest point) algorithm. M-estimator is applied in the ICP algorithm to improve the robustness to noises. Odometry data and UKF (unscented Kalman filter) are utilized to address the local minimum problem in ICP and to improve the accuracy. Experimental results with both synthetic and real data prove the reliability and high accuracy of the proposed approach.