3D reconstruction and motion estimation using forward looking sonar
Abstract
Autonomous Underwater Vehicles (AUVs) are increasingly used in different domains
including archaeology, oil and gas industry, coral reef monitoring, harbour’s security,
and mine countermeasure missions. As electromagnetic signals do not penetrate
underwater environment, GPS signals cannot be used for AUV navigation, and optical
cameras have very short range underwater which limits their use in most underwater
environments.
Motion estimation for AUVs is a critical requirement for successful vehicle recovery
and meaningful data collection. Classical inertial sensors, usually used for AUV motion
estimation, suffer from large drift error. On the other hand, accurate inertial sensors are
very expensive which limits their deployment to costly AUVs. Furthermore, acoustic
positioning systems (APS) used for AUV navigation require costly installation and
calibration. Moreover, they have poor performance in terms of the inferred resolution.
Underwater 3D imaging is another challenge in AUV industry as 3D information is
increasingly demanded to accomplish different AUV missions. Different systems have
been proposed for underwater 3D imaging, such as planar-array sonar and T-configured
3D sonar. While the former features good resolution in general, it is very expensive and
requires huge computational power, the later is cheaper implementation but requires
long time for full 3D scan even in short ranges.
In this thesis, we aim to tackle AUV motion estimation and underwater 3D imaging by
proposing relatively affordable methodologies and study different parameters affecting
their performance. We introduce a new motion estimation framework for AUVs which
relies on the successive acoustic images to infer AUV ego-motion. Also, we propose an
Acoustic Stereo Imaging (ASI) system for underwater 3D reconstruction based on
forward looking sonars; the proposed system features cheaper implementation than
planar array sonars and solves the delay problem in T configured 3D sonars.