Closing the Loop: Pursuing a Moving Object by a Moving Observer

Peter Nordlund, Tomas Uhlin

Tech. Rep. ISRN KTH/NA/P--95/08--SE, March 1995.


We present an integrated system, able to pursue a moving object by controlling a robot-head, so that the moving object is maintained in the center of the image. This system runs continuously in time and updates the object localization at a frame-rate of 25 Hz. The moving object can be tracked although the observer performs an unknown independent motion, involving both translation and rotation. We focus on a simple algorithm, since computational cost is of major importance when it comes to real-time systems with feedback. The method exnadats the brightness constancy constraint in conjunction with an affine fit between 2 consecutive images. After the affine fit has been made, an error/residual image is produced. A centroid of this error image is computed. The centroid indicates the location of the object. The algorithm is noniterative and computationally cheap. The running time complexity is O(n) for an image containing n pixels. Currently the real-time implementation computes only the translational component of the affine fit, whereas the full affine fit can only be performed offline. The image-processing takes place on a MaxVideo200 pipeline processor. The head control algorithm is running on a Transputer T800 network. Also some offline experimental results are presented.

Keywords: normal velocity, real-time system, tracking

Tech report: (PostScript 1.9M) CAIP'95 version: (PostScript 690k)

Peter Nordlund <>

Tomas Uhlin <>