Unscented Kalman Filter-based Video Object Tracking

Wide range of applications With the rapid development of digital image technology and digital media, video-based moving target detection and tracking has become an important part of the digital video technology, it not only requires the ability to phase in a video will be detected targets with video backgroundsteps to lay the foundation for separation, but also for the subsequent target tracking, and trajectory analysis. Moving target detection of objects detected in the video image sequence of image motion relative to the whole scenes, video target tracking technology has been widely used in security monitoring, regional monitoring, traffic management, vehicle, military control, and live entertainment, and many other occasions, has become a hot topic in the field of vision research.Target tracking one of the most difficult tasks in computer vision when this difficulty is mainly from changes in shape, camera and target motion and objectives of the context in which caused, in particular, the deformation of the target, it may be with the affine change, including position, rotation, shear deformation or scale changes. In addition, the target may be composed of several hinge body, in this case, the deformation of the detection target may even be completely flexible. In this case, the use of B-spline curve can effectively avoid the error in the detection and tracking of moving targets. In addition, the estimated and real-time tracking problem in the state of the video images of moving targets, often use the Kalman filtering method. Used in this paper is a combination of the Unscented kalman filter algorithm and B-spline strip curve target tracking method effectively overcome the problems of the conventional method in nonlinear problems and complex background target easily dissipate. Simulation show that Unscented kalman filter algorithm and B-spline strip curve can executce the objectives of the video motion extraction and detection, and thus lay the foundation for the next step of tracking and analysis.This article is the background subtraction method, snake model, unscented Kalman filter method combined, used in the video moving target tracking problem, a detailed analysis of the stability of the existing video object tracking method, denoising, real-time and positioning accuracy of the target, on the basis of various algorithms the advantages and disadvantages of the introduction of the B-spline curve tracing method used to achieve the contour of the moving target detection and recognition. The detailed description of the tracking algorithm based on unscented Kalman filter method and B-spline curve moving target strip and details of the process, and the simulation experiments, comparative analysis of the advantages and disadvantages of this method with other algorithms, and The simulation results of the error analysis. Experimental results show that based on the unscented Kalman filter Law and the B-spline curve moving target tracking can effectively and visually detect the true trajectory of the moving target, and to achieve a prediction on this basis, the location of the target to meet the real-time requirements.