Abstract: The vehicle camera calibration is one of the key technologies of lane departure warning system(LDWS) based on machine vision. In according with perspective projection principle and pinhole imaging model, the external static parameters of the camera are obtained by using trilinear calibration method. Based on the characteristics of vehicle jolt during traveling, the influence factor of camera calibration precision is analyzed via the mathematical model. A dynamic compensation algorithm based on the road characteristics is proposed. The pitching angle and height of vehicle camera is dynamically compensated on the basis of the inherent characteristic that two lane lines of the road are parallel and the width is invariable. The lateral position of the vehicle in the lane and the yaw angle of vehicle running in the lane are calculated by using the pitch angle and height values after compensation. A test platform independent from the current vehicle camera system is designed and used to carry out dynamic test of the compensation algorithm. The results of simulation and real vehicle test show that the compensation algorithm can reduce the calibration error caused by vehicle jolt obviously, thus effectively reducing the measurement errors of lateral position and yaw angle of vehicle.

Key words: Camera calibration, Dynamic compensation, Lane departure warning system(LDWS), Vehicle active safety