IMU-Assisted Calibration of Visual Observation for Vehicle Lateral Positioning

This paper develops a vehicle lateral positioning method, which combines visual measurement with inertial measurement unit (IMU) information. Firstly, we extract and fit the lane line with a lightweight lane detection algorithm. We then obtain the initial lateral distance between the vehicle and the adjacent lane line with a monocular camera. Next, we apply an IMU to correct the initial lateral distance by compensating for the camera relative pose. In this way, the negative effects of vehicle attitude changes on visual lateral positioning can be reduced. Therefore, this method can achieve better lateral positioning when the relative pose of the camera changes dramatically. Finally, experiments with real vehicles show the effectiveness of the proposed method.