Experimental Evaluation of Object Detection Algorithms for UAV Tracking

  • Author / Creator
  • Object detection is an image processing technology to detection different classes of objects using computer vision, i.e. putting bounding boxes over objects from a camera video feed. A landmark detection method was the Viola-Jones Algorithm introduced in 2001. The object classifier in this algorithm is called the Cascade Classifier, and was used to detect human faces in real-time. However, the Cascade Classifier is poor at detecting objects which have different features or poses than on those it was trained, or when the object is presented with a complex background. In order to achieve a more robust and accurate detection result than Cascade Classifier, deep learning techniques are used as an object detection solution for real-time aerial robotics detection. Two state-of-the-art deep learning frameworks, Darknet and TensorFlow, were implemented in Robot Operating System (ROS) and tested in experiment. Computational costs and reliability among various classifiers in the two deep learning frameworks were tested and compared. The experiments were conducted on a commercially available Parrot AR.Drone 2.0 Unmanned Aerial Vehicle (UAV) flying over various backgrounds. Experimental results show that a real-time, accurate and consistent UAV detection can be achieved by using deep learning techniques, and the results can be extended to the more challenging case of UAV tracking.

  • Subjects / Keywords
  • Graduation date
    Fall 2019
  • Type of Item
  • Degree
    Master of Science
  • DOI
  • License
    Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. Where the thesis is converted to, or otherwise made available in digital form, the University of Alberta will advise potential users of the thesis of these terms. The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission.