Anderson, James R. (2013) Stereovision in underwater environments. Masters thesis, Memorial University of Newfoundland.
- Accepted Version
Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission.
Stereoscopic vision is the modern field of using multiple cameras to extract three dimensional information about a scene. This technology is used in a wide variety of applications from motion capture used in the movie industry to the industrial monitoring and validation of production lines. This technology however has seen limited use in the challenging environment of underwater photography. This thesis attempts to implement and adapt this technology for use in the Marine Institute flume tank. The flume tank is used for scientific modeling and validation of fishing gear and other objects in ocean environments. This works focuses on the challenges involved in doing this, as well as experimental validation of modern camera calibration and triangulation and adding several novel improvements on these processes. -- This works shows that a modern system using a properly calibrated system functions faster, more accurately and more precisely than any human driven monitoring system. The testing of the various modern calibration techniques reveals several weaknesses when exposed to the challenging underwater environment. The comparison of several methods for stereo location showed the accuracy of these methods is greatly reduced in challenging environments. Both these results open the way for several novel improvements on the methods which increase accuracy and improve performance over the original methods.
|Item Type:||Thesis (Masters)|
|Additional Information:||Includes bibliographical references (leaves 149-153)|
|Department(s):||Engineering and Applied Science, Faculty of|
|Library of Congress Subject Heading:||Photography, Stereoscopic; Underwater imaging systems; Computer vision.|
Actions (login required)