Research On Tracking Algorithm Based On Multi-feature Fusion And Multi-template

The visual object tracking algorithm is mainly to study a given video sequence.Firstly,the position and size of the target in the first frame of video will be marked;secondly,the algorithm will be used to predict the position of target;finally,obtain the target's trajectory.This technology provides an important basis and foundation for the analysis and understanding of video sequences,and has a wide range of applications in life.Military aspects include unmanned aerial vehicles(UAV),precision guidance(PGW),airborne early warning(AEW),battlefield surveillance(BSL),etc.;civilian aspects include mobile robots((AMR),intelligent video surveillance(IVSS),intelligent transportation systems(IVHS),Human-Computer Interaction System(HCI)and Virtual Reality System(VR),etc.Although object tracking has achieved good results in the past research process,when the tracking algorithm is applied to the actual environment,the effect is still not satisfactory.Therefore,the object tracking algorithm still faces huge challenges in practical applications.Mainly include on these aspects: The target scale changes drastically,the target is occluded or even completely occluded,the motion blur caused by the target movement too fast,the target's violent deformation,single object tracking and multiple object tracking,etc.Based on the analysis of the problems that still exist in the above-mentioned object tracking field,this paper will improve the basic algorithm BACF through methods such as feature fusion,response value peak discrimination and multi-template update.The algorithm in this paper is compared with a variety of excellent object tracking algorithms in recent years based on OTB100.The algorithm in this paper has better tracking performance in complex environments such as partial occlusion,severe deformation,motion blur,and background chaos.It means that the algorithm in this paper is more adaptable to various complex tracking scenarios than other comparison algorithms,and it can show excellent accuracy and robustness.Especially the success rate and accuracy value are greatly improved compared with the basic algorithm of this paper.