Index of /dsl-cdrom/unsw/projects/track

 Name                    Last modified      Size  Description
 Parent Directory                             -   
 bccsrc/                 29-Nov-2004 17:14    -   
 hccsrc/                 29-Nov-2004 17:14    -   
 poster/                 29-Nov-2004 17:14    -   
 report/                 30-Nov-2004 10:44    -   
 resources/              29-Nov-2004 17:14    -

Companion CD-ROM featuretracking Directory

Current Directory:	CD-ROM Root / unsw / projects / featuretracking • View as a list of files.
Quick Links:	CD-ROM Root • unsw • gnutools • xilinx • examples • reference • board • utils

Companion CD-ROM: /unsw/projects/featuretracking directory

The /unsw/projects/featuretracking directory contains the source code, documentation and resources used in this project.

This project was done as part of a final year thesis by Ooi Yi Tung in 2004.

Video Feature Tracking Device

Introduction

This thesis is about designing a hardware implementation of a video feature tracking device. It is split into a few parts. First, a software is designed to simulate the working environment on hardware. The aim of the software is to familiarize myself with the processing of video frames and the identification of unique features from a live video stream. Tracking algorithms are also designed, calibrated and tested using this software, since recompilation of source code and rebuild of software take a much shorter time. After that, the feature tracking device will be redesigned and optimized on a development board, the Celoxica RC200 Expert Kit. The final design is Field Programmable Gate Array (FPGA)–based and requires additional hardware support such as a web camera. Finally, the hardware and the software are integrated via a standard communication link, the parallel port, where the software is able to analyze information obtained by the hardware.

This thesis is particularly interesting since feature tracking devices are mostly designed with software. High level design in software is undoubtedly productive, but not necessarily efficient. Some algorithms that demand heavy processing just cannot be implemented in software due to the limited processing power available to user applications, the operating system itself consumes a huge part of the processing time. Furthermore, in a tracking device, real-time processing is sometimes necessary. Precise timing is required for high accuracy analysis. This can be done effectively in low level hardware design, where the designer has full control down to a single clock cycle.