Target tracking and range estimation using an image sequence

Abstract

Describes a new method of tracking targets across a sequence of images and estimating the range of these tracked objects from the camera. The motion of the camera between the successive frames is assumed to be known. The distance from the sensor to the objects is assumed to be much larger than the physical dimensions of the object. the approach presented makes use of the known sensor motion to generate expected images which are then used in establishing a reliable correspondence and in tracking the targets across the image sequence. A mean normalized area correlation method is used to establish the inter-frame correspondence. Once this correspondence is established, two new methods of estimating the range of the target from the sensor are also developed. The techniques developed are tested with good success rate on a sequence of real aerial images.<>