Date of Award

3-22-2019

Document Type

Thesis

Degree Name

Master of Science in Computer Science

Department

Department of Electrical and Computer Engineering

First Advisor

Scott L. Nykl, PhD

Abstract

Aerial real-time surveillance exists in a paradigm balancing the constraints of delivering high quality data and transporting data quickly. Typically, to have more of one, sacrifices must be made to the other. This is true of the environment in which an Unmanned Aerial Vehicle (UAV) operates, where real-time communication may be done through a low-bandwidth satellite connection resulting in low-resolution data, and serves as the primary limiting factor in all intelligence operations. Through the use of efficient computer vision techniques, we propose a new Structure from Motion (SfM) method capable of compressing high-resolution data, and delivering that data in real-time. Specifically demonstrating a 90 percent compression of original video imagery at 4 Hz which equates to an 80x computation time speed-up compared to traditional SfM methods, with an added benefit of presenting the original 2D intelligence data as a 3D virtual model

AFIT Designator

AFIT-ENG-MS-19-M-007

DTIC Accession Number

AD1074014

Download

Share

COinS