Research On Visual Tracking Algorithm Based On Spatio-temporal Context

Visual tracking is the analyses,processes and understandings of the useful information in the video sequence to estimate the target position in each frame,and to complete the tracking task to help people better perceive the real world.Although the visual tracking technology has made great progress in recent years,the design of a visual tracking method in complex scenes is still a very challenging topic due to a disturbance such as illumination changes,scale changes,occlusions of all kinds of factors.Based on the studies of spatio-temporal context and visual tracking theory.The method of spatio-temporal context target object tracking in complex scenes is researched,and an improved strategy is proposed in this paper.Contents are as follows:1 Through the study of spatio-temporal context target tracking method,we can get the relevance between the spatio-temporal context by modeling the temporal relationship of visual tracking and local context regions.Then assess the position of the new target in a frame throughout the relationship between time and space and biological vision system.The simulation results show that the proposed algorithm can run in the speed of 350 frames per second and have good real-time property,because training and recognition are realized by fast Fourier transform.However,the algorithm is updated errors easily when the target scale changes and occluded.2 Aimed at the problem of when the target object scale changes and occluded is updated errors,which leads to the problem of tracking drift in the spatio-temporal context of visual tracking algorithm,mistakes are made easily.A kind of multi-scale context target object tracking algorithm is proposed in this paper.The simulation results show that this proposed algorithm can complete the robust tracking in the position of scale changes,local occlusion and pose variations.The tracking success rate in the proposed algorithm is improved by 38.4% than the tracking success rate in original spatio-temporal context tracking algorithm.Experimental results show that the proposed algorithm has some advantages in real-time property and robustness.