Multi-Sensor Fusion for Lateral Vehicle Localization in Tunnels

The satellite navigation signal in the tunnel is weak, and it difficult to achieve accurate lateral positioning complex conditions such as low-speed congestion by relying solely on inertial or line image recognition, which one of problems automatic driving at present. In this paper, a lane-level location method based multi-sensor fusion proposed. Using machine vision method, detecting lane lines with monocular camera, fitting determine status vehicle information. top view taken binocular distance from width are calculated pictures camera. Obtaining heading angle information using gyroscope odometer. When car changes lanes overtakes, new calculating difference combining odometer so complete vehicle. simulation results show that algorithm has high accuracy. accuracy less affected drift elements, error will not accumulate.