Design and Implementation
of a
Real Time Object Tracking System
A thesis submitted in partial fulfillment of
the requirements for the degree of

Bachelor of Technology
in
Computer Science and Engineering

Amit Aggarwal
Lalit Kumar
under the guidance
of
Dr Subhashis Banerjee


Department of Computer Science and Engineering
Indian Institute of Technology, Delhi
May 1997

CERTIFICATE

This is to certify that the thesis entitled Design and Implementation of a Real Time Object Tracking System, being submitted by  Amit Aggarwal and  Lalit Kumar, is a record of bona fide work conducted by them under my supervision.

The matter and results embodied in this thesis are original and have not been submitted elsewhere, wholly or in part, for the award of any degree or diploma.

Dr. S. Banerjee
Department of Computer Science and Engineering
Indian Institute of Technology
Delhi

ACKNOWLEDGMENTS

We are deeply indebted to our supervisor, Dr. S. Banerjee for the continuous encouragement, guidance and support he provided us during the course of this project. He was always there to help us in times of need. It would not be an exaggeration to say that without his technical advice and help, this project would not have been a reality.

We are also thankful to Pankaj Jain and Ashish Thusoo, who made us familiar with the key concepts involved in object tracking. Their ideas have formed the basis of this project.

Thanks are also due to Mr. Kaushik for the cooperation we received from him regarding the facilities in the Artificial Intelligence laboratory.

Amit Aggarwal Lalit Kumar

Abstract:

This thesis describes the design and implementation of a real time object tracking system. Traditional approaches to tracking have been based purely on intensity correlation matching with no attempt made to impose structural constraints on the features being tracked. This report presents a new structure based tracking scheme, valid under the simple assumptions that the camera model is affine and the object being tracked is undergoing affine motion. The approach attempts to unify tracking and structure extraction, such that each of them aids the other. Correlation matching is used to compute the object structure while the quality of matches is improved by imposing structural constraints. The algorithm has been implemented in real time using a Yorick robot head. Results demonstrate that the algorithm is fast, reliable, view point invariant, insensitive to occlusion and/or disappearance of individual corners, and provides a smooth gaze demand.

• Contents
• List of Figures
  • Introduction
    • Motivation
    • Our Approach
    • Objectives and Goals
    • Structure of the Report
• Tracking Algorithm
  • Theoretical Background
    • Kalman Filtering
    • Affine Multiple Views Geometry
  • The Tracking Algorithm
    • Basic Algorithm
    • Modules
    • Alternatives and Analysis
    • Final Algorithm
  • Off-line Simulation and Analysis
    • Offline Implementation
    • Summary of Results and Conclusions
• Real Time Implementation
  • Hardware and Software Support
    • Hardware
    • Software Libraries
    • Overall Setup
  • Implementation
    • Modules
    • Overall Implementation
  • Interfacing with Yorick
    • The PMAC Card
    • Stability of Closed Loop Control
    • Spot Tracking Experiment and Choice of Control Strategy
  • Conclusion
    • Results of Real Time Tracking
    • Features of the System
    • Future Work and Applications
  • XIL Molecules
    • Deferred Execution
    • XIL Molecules
    • Molecules used in the Implementation
  • Control Programs for PMAC Card
    • PVT Mode Moves
    • PMAC Program for Tracking Application
• References
• About this document ...