Award Date

2009

Degree Type

Thesis

Degree Name

Master of Science in Electrical Engineering (MSEE)

Department

Electrical and Computer Engineering

Advisor 1

Emma E. Regentova, Committee Chair

First Committee Member

Muthukumar Venkatesan

Second Committee Member

Henry Selvaraj

Graduate Faculty Representative

Ajoy Kumar Datta

Number of Pages

185

Abstract

We describe a real time system for recognition and tracking 3D objects such as UAVs, airplanes, fighters with the optical sensor. Given a 2D image, the system has to perform background subtraction, recognize relative rotation, scale and translation of the object to sustain a prescribed topology of the fleet. In the thesis a comparative study of different algorithms and performance evaluation is carried out based on time and accuracy constraints. For background subtraction task we evaluate frame differencing, approximate median filter, mixture of Gaussians and propose classification based on neural network methods. For object detection we analyze the performance of invariant moments, scale invariant feature transform and affine scale invariant feature transform methods. Various tracking algorithms such as mean shift with variable and a fixed sized windows, scale invariant feature transform, Harris and fast full search based on fast fourier transform algorithms are evaluated. We develop an algorithm for the relative rotations and the scale change calculation based on Zernike moments. Based on the design criteria the selection is made for on-board implementation. The candidate techniques have been implemented on the Texas Instrument TMS320DM642 EVM board. It is shown in the thesis that 14 frames per second can be processed; that supports the real time implementation of the tracking system under reasonable accuracy limits.

Keywords

Computer recognition; Flying objects; Real time; Three dimensional object recognition; Tracking objects; Two dimensional images

Disciplines

Computer Engineering | Electrical and Computer Engineering | Military and Veterans Studies

Language

English


Share

COinS