Research On Visual Servoing-based Autonomous Vehicular Follower System Under Complicated Environments

This paper focuses on the functional parts of the autonomous vehicular follower systems.Visual tracking based on visual servoing has stable performance when equipped with pan-tilt union.And trajectory planning based on the base frame makes the vehicle follow the leader autonomously.The main achievements are as follows:1.We proposed color name TLD.The introducing of color name feature improves the performance of TLD substantially.We tried many kinds of features to improve the performance,concluding that color name is the best.The proposed method was tested in three kinds of analysis modes on 20 datasets under different scenes.Experimental results show that the stability and accuracy outperform other publicly available tracking algorithms.2.We studied on feedback and visual servoing method based on Kalman filter.Kalman filter was employed to estimate the motion station.On the one hand,to compensate the delay of visual tracking under complicated environments,we sent the motion estimation to pan-tilt controller.On the other hand,we proposed the self-adaptive feedback searching method based on Kalman filter,which improves the search efficiency of tracking algorithm.3.We improved curvilinear coordinate transformation trajectory planning framework.In Trajectory planning based on arc joint,look-ahead point is set as current target position without consideration of the terminal heading angle which lead to the terminal heading angle uncontrollable.To solve this problem,this thesis improved the curvilinear coordinate transformation trajectory planning framework by introducing a relaxed terminal constraint.Then a number of trajectories can be acquired,from which the best can be chosen in the next stage.