Research On Key Technology Of Template Update In Visual Object Tracking

Visual object tracking technology is one of the main components in the field of computer vision.This technology has been widely used in areas such as smart security,precision guidance,and human-computer interaction.The task of object tracking is to estimate the position,scale and other information of the tracked object in the image frame with continuous video image sequence,and determine the moving speed and direction of the object to adapt more advanced computer vision tasks.In the process of object tracking,it is easy to cause the tracking template to drift when the tracked object is affected by factors such as illumination changes,occlusion,and object blur.Tracking template is the key of tracking algorithms to track accurately as we know.Therefore,we do some researches on some key technologies of template update in the visual object tracking algorithm.In the field of object tracking based on correlation filtering,the template update methods with traditional object tracking algorithms are unsatisfactory when faced with complex environments.In order to improve the robustness of the tracker and reduce the probability of drifting in tracking template in the trackers,we designed a novel incremental multi-template update strategy.This strategy analyses that different historical filter templates have different contributions to the tracker to achieve accurate tracking,and shows that a highly reliable filter template is the key to track accurately.Therefore,in the tracker template update process,we incrementally merge the filter template with the highest reliability in the local history with the tracker template.This template update method is significantly different from the template update methods of traditional correlation filter tracker algorithms.This method can effectively alleviate the influence of many interference factors on the trackers.The experimental results of three test benchmarks,including VOT2016,OTB100 and UAV123 datasets,show that the performance of our trackers is superior to that of the state-of-the-art trackers.In the fast-moving state of the object,the possibility of the change of appearance and the object being blurred increases,so it is very easy to cause the trackers to miss object.In order to reduce the impact of the object's rapid movement on the performance of the tracker,we designed a template sparse update tracker based on the object's moving speed.In our proposed template update interval adaptive tracker,the object's moving speed is judged according to the change of the maximum response point of the object in several consecutive frames,and the template update interval frame number and scale change factor are adaptively adjusted according to the object's moving speed.The comparative test on the VOT2016 data set proves that the template sparse update tracker based on the object moving speed that we proposed has a strong ability to cope with the fast moving of the object.