Vision-Based Technique to Estimate the Roll Angle of Powered Two-Wheeled Vehicles

This paper presents an innovative vision-based method for estimating the roll angle of Powered Two-Wheeled Vehicles with respect to the road surface. This method uses a single camera attached to the vehicle as an input. The algorithm is based on a novel technique that exploits perspective distortion and Inverse Perspective Mapping to calculate the sub-pixel distance difference between the edge and center lane markers and estimates the camera roll angle. The vehicle roll angle is then obtained in the Euler convention. This method is tested on various scenarios with different speeds and road shapes using BikeSim simulator. The results are promising, showing a low estimation error. Moreover, its main advantage is that the estimation is not cumulative, thus avoiding drift errors over time.