Research On Anti-Interference Single Object Tracking Algorithm In Complex Scene

Object tracking is a very important branch of computer vision,its main purpose is to find the position and size of the target in the image.The primary task of target tracking is to specify the position and size of the target in the first image frame and predict the position and size of the target in the subsequent video image frame.Nowadays,single target tracking in computer vision has become a hot and main research direction in the field of image processing.Many scholars at home and abroad have done a lot of research on target tracking task,and achieved good tracking effect in many simple scenes.However,it is difficult to achieve accurate and effective target tracking in complex realistic scenes.In this paper,the following two aspects of work to improve the performance and efficiency of target tracking:First,the scene of object tracking is very complex,and there will be various challenges and interference factors in the process of target tracking.Therefore,we design a tracker that can resist interference and challenge.Most existing trackers use a single feature or fixed fusion weights,which makes tracking likely to fail in the case of deformation or severe occlusion.Therefore,this paper proposes an adaptive fusion strategy of multi-feature response graph based on the consistency of single feature and fusion feature.The performance of the whole tracking system can be improved by accentuating the appropriate feature response of the current scene and reducing the inappropriate feature response,which greatly improves the robustness and accuracy.In addition,when the target is occluded,the response graph obtained after the relevant operation has multiple local peaks,so we propose an anti-occluding mechanism.Specifically,if the non-maximum local peak meets certain conditions,we will move the search area and retest the location and size of the target to obtain a new response graph.Then,we select the response graph with the maximum response value as the final response graph.Experimental results show that this anti-occlusion mechanism can effectively deal with tracking failure scenarios caused by occlusion.Finally,a model updating strategy with high confidence is designed to deal with model contamination by adjusting the learning rate in different scenarios.Excellent results of 90.71%and 68.26%in accuracy and success rate were achieved on the OTB2015 dataset,respectively,and the same has a very large advantage in comparison with advanced tracking algorithms on UAV123 and TC128 target tracking data.Second,UAV target tracking is one of the most active in the field of application,remote sensing in unmanned aerial vehicle(uav)Angle tracking scenarios,the movement of the target faster than ordinary scene,when the unmanned aerial vehicle(uav)rapid movement,may lead to the target in the next frame to the target search area edge or directly outside the target search area,makes the tracking accuracy decline or even failure of the algorithm.In addition,when the target is disturbed by similar objects or partially blocked,the existing tracker is also very easy to lose the target.Aiming at the shortcomings of the above existing technologies,this paper designs an anti-jamming single target tracking algorithm for UAV.On the one hand,the target is partial occlusion interference or similar goals lead to failure of the problem,this paper proposes a regularization correlation filter based on response bias awareness,when the target is obscured or similar object interference,response figure will appear many local wave interference peaks,such as by inhibiting the partial response figure interference wave position,to smooth the response to a change,Enhance the tracker’s perception of target appearance changes and improve the robustness of the tracking method to deal with different scenarios.On the other hand,in order to overcome the uav to track moving targets in the scene too fast lead to tracking failure problems,based on kalman filter is proposed in this paper a kind of search method of center of interest,by using kalman filter to determine the destination of interested in center position,so it can greatly enhance the accuracy of the tracking algorithm.The excellent performance of 69.4%and 48.5%on UAV tracking dataset UAV123 in terms of accuracy and success rate,respectively,also has a large advantage in comparing with advanced tracking algorithms on UAV target tracking data of different scenarios such as UAVDT and DTB70.