Based On Partial Differential Equations Of The Dynamic Target Tracking Algorithm Design And Application

Object tracking based on the image series is the basic problem in computer vision. It can be used in the many directions includes video surveillance , medical image analysis , motion reconstruction and so on. In the field, the popular methods include: using Kalman filter or Particle filter to estimate motion, estimate the optical flow, utilizing the active contour model, the active shape model and the deformable template to tracking the objects, and using the Geodesic active contour model or Geodesic active region model which based the partial difference equation method and level set frame.Firstly, we summarize the basic methods and models of object tracking. We study mostly the PDE based the active contour model in recent years such as the level set frame , Geodesic active contour model and Mumford-Shah model.We brought out a new tracking method combining the traditional optical flow estimation and the tracking method based on the level set frame. The method also considered the probability of the object, the boundary energy of the object and the length energy of the contour. We also designed a new fast scan algorithm for the method. The method can reduce the computer complex and the iterative number. The result of the experiment shows the method can exactly track the selected objects.