Research And Application Of Target Tracking Algorithms Based On Compressive Sensing

Target tracking algorithm has been widely used in medical diagnosis,human-computer interaction,intelligent monitoring systems and many other fields,and it is one of the hot spots in the field of computer vision.Target tracking algorithm based on compressive sensing theory has been widely concerned by scholars at home and abroad because of its high tracking speed and accuracy.However,due to using a fixed tracking scale,both CT and FCT algorithms are prone to generate tracking drift or even lose the target when the target's size changes.To overcome the drawback,this paper proposes a method to estimate the size of target,and then proposes an improved compressive tracking algorithm.At the same time,the tracking algorithm need to detect the target and track it automatically in actual scene(e.g.intelligent video surveillance system),because the initial target position in tracking algorithm is usually manual selection.Therefore,this paper also makes a research on the detection algorithm,and finds a kind of target detection algorithm which is suitable for actual scene.The main works in this paper are as follows:(1)This paper introduces the research background,significance,and status at home and abroad of the target tracking algorithm,and summarizes the principles,advantages and disadvantages for several classical target tracking algorithms.This paper introduces several classical compressive tracking algorithms,and analyzes their advantages and disadvantages.(2)In order to solve the defect that the compressive tracking algorithm can't track target with different scales,this paper studies and proposes an improved compressive tracking algorithm to adapt the variable of target scale.Experimental results show that the proposed algorithm can achieve the goal of adaptive target scale tracking,and the average tracking success rate is 7.3 percentage points higher than that of the FCT algorithm.(3)In order to ensure the accuracy of the scale estimation,the back scale estimation is added.On the basis of this estimate,a compressive tracking algorithm with back scale estimation(BSCT)is proposed.Experimental results show that BSCT adapts well for the change of target's size and also performs well in reducing the influence of such as occlusion,deformation.rapid movement and background clutter.Moreover,BSCT is capable of real-time tracking with higher robustness,accuracy.(4)I realize the ASCT algorithm and BSCT algorithm described above using VS2012,and do experiments compare with several tracking algorithms,which verifies the effectiveness of these algorithms.(5)Because the actual scene often need to detect the target,and then carry on the target tracking,so I research and realize a target detection algorithm based on background subtraction and three frame difference using the VS2012 platform.At the same time,the algorithm is applied in the actual scene,and the results are compared with the experimental results of the background subtraction and the three frame difference algorithm.The experimental results show that this algorithm has better adaptability to the environment and can better deal with the jitter in the scene,and the detection of the target area is more complete.