Research On Tracking Technology Of Vehicle Under Complex Ground Background

Vehicle tracking technology is of great significance in practical application.It is an important part of ITS.However,due to the variable movement of the vehicle and the disturbance of the external environment,there are many challenging problems to solve in vehicle tracking.The studies are carried out under the scenario that the vehicles are under complex ground.The influence of turning,partial or total occlusion,illumination variation,complex background and camera movement on the tracking are studied.The algorithm should be adapted to these disturbances,and runs at a fast tracking speed.The specific research contents of this paper are as follows:1.Tracking algorithm needs target's initial position and size before tracking.In this paper,a fast segmentation and extraction method of target under complex ground is proposed.According to the boundary discontinuity of gray between the target and background,the algorithm extracts the edge of vehicle,combines with the morphological processing technique,and statistics connectivity domain to screen target.Experimental results show that the algorithm can quickly segment and extracts the target.2.STC uses the grayscale feature,which is insufficient in the description of the target appearance when deformation,rotation,or blur occurs.For this problem,a Spation-Temproal Context tracker based on Color information is proposed,which performs better than STC in deformation,in-plane rotation,out-of plane rotation,scale and so on.3.Aiming at solving the deficiency of the above algorithm and meeting the demand of practical application,a real-time vehicle tracking algorithm based on edge information is proposed.The algorithm adds the edge feature to the prior model.Besides,the algorithm proposes a mechanism of judging tracking failure and redetection target after tracking failure.What's more,the algorithm improves the model updating strategy.Finally,numerous experiments are conducted to prove that the algorithm can basically solve the problems of illumination variation,vehicle turning,camera movement,occlusion,redetection and so on during the vehicle tracking process.4.The vehicle segmentation and tracking algorithm are implemented on the DSP TMS320C6455 and optimized.After optimization,the tracking speed of the system can reach50 frames / s.Finally,in field experiments,the vehicles in the actual ground scenes are tracked,which proves the robustness of the algorithm and the real-time performance of the system.