Colour-based Object Tracking in Surveillance Application

The motivation of the proposed method is to resolve typical tracking's challenge which is object's occlusion in the scene. In this paper, we proposed a method to track the objects consistently in real-time unconstraint surveillance application. The proposed method capable to detect moving objects, track the objects appear in the scene and provide consistent identifier for tracked objects. We describe the characteristic of the motion tracker which based on colour as the key feature to compare the object's similarity. Initially, we produce a motion map which delineates the foreground and background. Then, the motion map is segmented and analyzed in order to remove noise and connect into motion regions. Subsequently, motion tracking is performed based on colour-based feature matching. The colour information is extracted cluster-by-cluster to compare the object's similarity across the image sequences. We ensure the efficiency of the method with assembly an assumption that only entire objects are allowed to be track. In order to support the assumption, we utilized wide-view camera to avoid capturing partial objects.