Analysis and Compensation of Acoustic Rolling Shutter Effect of Acoustic-Lens-Based Forward-Looking Sonar

Abstract

As the demand for ocean observations increases, the quality requirements for sonar imaging are becoming increasingly exacting. The temporal synthesis strategy provides the acoustic-lens-based forward-looking sonar (ALFLS) with ultrahigh resolution but also causes the imaging content to be distorted when the sonar is moving at a high speed. This limits the application of sonar to low-speed or stationary underwater inspections. In this article, we formally define this distortion effect for the first time as an acoustic rolling shutter (ARS) effect and use implicit neural representation to naturally recover and represent acoustic images in a rectangular coordinate system for visual representation or subsequent computer tasks, e.g., image registration and image mosaicing. The method is self-supervised and involves only a single image, without the need for any external data input, e.g., from a global positioning system and a Doppler velocity log. We validate the effectiveness of the proposed method using experimental field data and reveal the potential application of this technique to underwater robot vision.