Adaptive Scale Appearance Model for Object Tracking

To track objects with scale changes in a complex scene,a particle filter based object tracking approach is proposed in this paper.The approach partitions the reference object into several sub-regions by clustering in color space,then the color distribution of each sub-region is modeled as Gaussian,and its location constitutes the spatial constrain on the layout of the object.The reference model is built from he color distribution and spatial relation of the sub-regions.The reference model is integrated into the particle filtering to search for the object location and detect the scale change of the object.The reliability of detecting the scale change of the object in a complex scene is improved thanks to the usage of both spatial and color information.Experimental results show that the proposed tracking approach is robust and effective to scale changes,pose variance and partial occlusions.