Side-attack explosive hazard detection in voxel-space radar using signal processing and convolutional neural networks

Author

Blake Brockner

Issuing Body

Mississippi State University

Advisor

Ball, John E.

Committee Member

Anderson, Derek T

Committee Member

Archibald, Christopher

Committee Member

Fowler, Jim

Date of Degree

8-9-2019

Original embargo terms

Worldwide

Document Type

Graduate Thesis - Open Access

Major

Computer Engineering

Degree Name

Master of Science

College

James Worth Bagley College of Engineering

Department

Department of Electrical and Computer Engineering

Abstract

The development of a computer vision algorithm for use with 3D voxel space radar imagery is observed in this thesis. The goal is to detect explosive hazards present in 3D synthetic aperture radar (SAR) image data. The algorithm consists of three primary stages; a precreener to find areas of interest, clustering for labeling distinct areas, and a classifier. The performance between multiple prescreener methods are compared when using a heuristic classifier. Finally, a convolutional neural network (CNN) is used as a classifier stage and a comparison between a deep network, a shallow network, and human experts is conducted.

URI

https://hdl.handle.net/11668/14590

Comments

voxel-space radar||side attack explosive detection||matched filter||size contrast filter||deep learning||convolutional neural networks

Recommended Citation

Brockner, Blake, "Side-attack explosive hazard detection in voxel-space radar using signal processing and convolutional neural networks" (2019). Theses and Dissertations. 3942.
https://scholarsjunction.msstate.edu/td/3942