A Novel Fusion Perception Method of Unmanned Driving Based on Distributed Roadside Information

Multi-sensor fusion, tracking, and prediction have a long history in signal processing, control theory, and robotics. This paper proposes a trajectory tracking method based on distributed cameras. For a start, we build an innovative model that calculates the observation noise for each camera. Secondly, we can identify the abnormal data by introducing the trustworthiness function. Then we use the weighted fusion method to fuse the data to minimize the variance. We will use the improved Bayesian filtering to deal with different observation noises for the final trajectory tracking. Simulation results confirm that the Root Mean Square Error (RMSE) of trajectory tracking using our method is reduced by at least 13% compared to those of single viewpoint perception. Even if several cameras are blocked, our approach can still perform best.