A Method for Object Tracking using Shape Matching

Using shape-initialized deformable template, a new method for object tracking is presented. This method consists of two steps: 1) searching step, by background predicting, moving object is segmented, then detecting the edge, thinning edge, and select landmark points to construct a shape model which contains all the feature points of the shape; 2) tracking step, using prior shape model which is extracted from the previous frames to implement the coarse location of object by shape matching, then adjusting shape model to obtain the inter-frame change of the object by adopting active contour model, to achieve accurate location and generate a updating shape model for object detection in next frame. The results of experiments demonstrate that the mentioned algorithm has strong adaptability to deformation, pose change, lighting change, and occlusion