Sonar Image Registration for Localization of an Underwater Vehicle

King, Peter and Anstey, Benjamin and Vardy, Andrew (2017) Sonar Image Registration for Localization of an Underwater Vehicle. Journal of Ocean Technology, 12 (3). pp. 68-90. ISSN 1718-3200

[img] [English] PDF
Download (664kB)

Official URL: http://www.thejot.net/

Abstract

This paper presents a system to provide augmented localization to an AUV equipped with a side scan sonar. Upon revisiting an area, from which side scan data had previously been collected, the system generates an estimate to bound the error in the AUV’s estimate. Localization is accomplished through the comparison of sonar images. Image comparison is based on the extraction of features which characterize local gradient distributions, such as Lowe’s SIFT feature extractor. To resolve potential ambiguities and noise in the image comparison measurement, the localization system incorporates a Bayesian inference algorithm that considers both image based measurement and relative motion to refine the position estimate over time. We describe the particular methods, constraints and augmentations used to apply established image matching and alignment techniques to side scan sonar imagery. By applying consistent geographical corrections to the raw sonar data; using a flat-bottom assumption; and by adding the constraint that images are formed with north aligned up; the traditional problem of full pose estimation is reduced to the two-dimensional case of determining only the x,y translation independent of vehicle altitude. Due to the assumption of constant scale and orientation between images, sensitivity of image feature matching is shown to be controllable by filtering feature matches based on comparing their scale and orientation. This effect was quantified using binary classification analysis. The system’s performance was measured by performing tests on a large side scan survey which represents the familiar terrain that a returning AUV could use for localization.

Item Type: Article
URI: http://research.library.mun.ca/id/eprint/13011
Item ID: 13011
Department(s): Science, Faculty of > Computer Science
Date: 2017
Date Type: Publication
Supplemental Date: 2017

Actions (login required)

View Item View Item

Downloads

Downloads per month over the past year

View more statistics