Detection-Verification-Cascade-Based Low-Cost Precise Vehicle Positioning Exploiting Extended Digital Map

This paper proposes a map-matching based precise positioning system for highways. The proposed fuses multiple sensors such as camera, conventional GPS, and vehicle motion widely adapted in mass-produced vehicles. While systems consume large amount of computation because the landmark detection, achieves both small computations by introducing detection-verification-cascade strategy: generates position candidates per lane utilizing endpoints detectable with computation, then it selects candidate on an ego-lane verifying existence road signs within region interest generated map each candidate. To reduce further, adapts extended Kalman filter (EKF) instead particle filter. As noise distribution EKF is limited to unimodal, maintains EKFs track until ego identified. After identification, leaves only one ego-lane. 0.17m average error highway situations takes about 53ms process whole procedure from detection estimation low-end hardware equipped cortex-A9 CPU whose clock 1.0 GHz.