A Color-Adaptive and Robust Visual Object TrackingMethod Based on MeanShift Algorithm

Visual object tracking is a key component in video analysis and surveillance system. In this paper we propose a novel and robust video object tracking method based on kernel tracking approach .MeanShift algorithm is a Kernel Tracking approach based on color histogram modeling. Because of changing of the color and shape of target model, it cannot track the object as much as possible. in some video streams with changeable color and brightness this method encounters with failure. So we manipulated some essential changes in original MeanShift and made it adaptive and more powerful in the realm of color, brightness and shape changes .The result of applying this method illustrates the high precision in our method for non-rigid objects in long videos.

کلیدواژه ها:

Kernel Tracking ، Color Weighted Histogram ، Target Model ، Target candidate