UACE2017 Proceedings: Estimating attitude and trajectory of forward looking imaging sonar using interframe registration

Session:
Signal and image processing
Paper:
Estimating attitude and trajectory of forward looking imaging sonar using interframe registration
Author(s):
Benjamin Henson, Yuriy Zakharov
Abstract:
The aim of this work is to produce an accurate attitude-trajectory estimate for a Forward Looking Sonar (FLS) based on the optical flow between consecutive sonar frames. The attitude-trajectory can be used to locate an underwater platform, such as an Autonomous Underwater Vehicle (AUV), to a degree of accuracy suitable for navigation.\n\nThe attitude-trajectory estimation is performed in three steps. Firstly, a selection of optical flow estimation techniques are applied to estimate a pixel displacement map between consecutive sonar frames in the native polar (range/bearing) format. From this displacement map, a reduced set of comparative statistics are produced.\n\nThe second step finds the best match for the displacement map statistics when compared to displacement maps for a set of modelled sonar sensor motions within an expected motion range. Thus, from the model, an estimate of the incremental sensor motion between frames is made.\n\nFinally, using a weighted regularised spline technique to integrate the incremental motion, an attitude-trajectory is generated for the sonar sensor.\n\nTo assess the accuracy of the attitude-trajectory estimate, the FLS frames are registered to a global Cartesian grid, building a mosaic of the underwater scene.
Download the full paper