Home // UBICOMM 2015, The Ninth International Conference on Mobile Ubiquitous Computing, Systems, Services and Technologies // View article

Object Location Estimation from a Single Flying Camera

Authors:
Insu Kim
KinChoong Yow

Keywords: computer vision; drone; stereo-vision; distance extraction; object detection; location mapping.

Abstract:
With the recent popularity and ubiquity of drones, there had been an increasing demand to deploy drones for the detection, localization and tracking of objects in a scene (e.g., pedestrians, cars, etc.). The problem with a single camera drone is that it is impossible to estimate distances from a single image. Although the drone can fly to another position to take a second image, the object that we are tracking may have moved during that time interval, rendering traditional stereo-vision algorithms useless. In this paper, we propose a novel system that instructs the drone to fly in a specific pattern so as to achieve a large baseline, and use three images (instead of the traditional two) to recover the distance to the object that is moving. The experimental results show that our algorithm can estimate depth with better or equal accuracy than other state-of-the-art methods. This algorithm would have great significance for small or low cost drones which are unable to carry additional devices (apart from the built-in camera), thus enhancing their ubiquity of use.

Pages: 82 to 88

Copyright: Copyright (c) IARIA, 2015

Publication date: July 19, 2015

Published in: conference

ISSN: 2308-4278

ISBN: 978-1-61208-418-3

Location: Nice, France

Dates: from July 19, 2015 to July 24, 2015