On the tracking of featureless objects with occlusion

Abstract

The author discusses an extremely efficient low-level tracking algorithm using only centers of gravity and sizes of the objects. He uses the center of gravity as the main tracking mechanism and the size of the object to aid in solving occlusion problems. Thus the size of the objects is mainly used as an aid in deciding how and if the number of objects changes from one frame to another. The author reports on experiments using video images of tennis balls bouncing through the field of vision. The camera is fixed and the lighting conditions are controlled. The balls are easily extracted from the background by subtraction of the image from a registered image of the background under the same lighting conditions. It is concluded that the proposed techniques can be applied to both video and infrared images, and are especially useful when the objects lack significant 'features' for matching.<>