Multiple Objects Tracking based on Auto-selection of Color Image Processing Methods

Abstract

In order to achieve the applications using vision sensors, it is impotrant to extract and track the moving objects from image sequences. Some tracking systems often prepare the models of the objects in advance. It is difficult to adopt this model based tracking in the general environment where many kinds of objects exist. On the other hand, model based tracking is efficient for the tracking in the crowded environment. We propose the color histogram based tracking system including the object model construction and object correspondence between frames by the fusion of the bottom-up process and top-down process.
In this paper, two processes that configure the tracking sysytem are described. Then, process selection method in the proposed tracking system is explained. Finally, experimental results of objects are shown and effect of the proposed tracking system is also explained.