Development of a Neural Network-Based Object Detection for Multirotor Target Tracking

Author

Spencer Harwood, South Dakota State UniversityFollow

Document Type

Thesis - Open Access

Award Date

2019

Degree Name

Master of Science (MS)

Department / School

Mechanical Engineering

First Advisor

Marco Ciarcià

Abstract

Unmanned aerial vehicles (UAVs) have, for the past few decades, seen an increased popularity in industry and research centres. Despite this intense utilization by both markets there exists an active demand for the development of autonomous guidance, navigation, and control strategies. One need relates to the achievement of a high level of autonomy to identify and track a target object. An elective technique for this set of tasks is neural networks. In the development and study of these networks there is a distinct lack of substantive validation techniques to qualify network performances when implemented in a multirotor UAV. This thesis will first describe the development of a neural network-based object detection subsystem for use in target tracking with an autonomous multirotor UAV. Then, the second part of this thesis will utilize a developed indoor multirotor testbed to externally verify the tracking performance of the multirotor UAV during an object following maneuver.

Library of Congress Subject Headings

Drone aircraft.
Drone aircraft -- Control systems.
Neural networks (Computer science)
Vehicles, Remotely piloted.

Format

application/pdf

Number of Pages

78

Publisher

South Dakota State University

Recommended Citation

Harwood, Spencer, "Development of a Neural Network-Based Object Detection for Multirotor Target Tracking" (2019). Electronic Theses and Dissertations. 3129.
https://openprairie.sdstate.edu/etd/3129