Wide Area Seabed Optical Imaging Using Autonomous Underwater Vehicles

Le résumé fourni par la source

Recent advances in Autonomous Underwater Vehicle (AUV) navigation, system reliability, and onboard data storage now enable efficient collection of high-resolution optical seabed imagery over survey areas spanning hundreds of square meters, achieving millimeter-scale pixel resolution. Integration of modern imaging systems with low-altitude AUV operations can capture tens of thousands of true-color images in a single mission, significantly enhancing capabilities for detailed seabed characterization, habitat mapping, and object identification. Advances with photogrammetric analysis allow for post-mission processing into wide area seafloor orthomosaics and digital elevation models. The integration of modern imaging technologies and offboard computing with advanced AUVs enables large-area underwater imaging for oceanography, habitat mapping, maritime archaeology, and marine debris. This paper presents efforts to integrate the Voyis Recon and Observer imaging payloads into the REMUS family of AUVs, specifically the REMUS 100, REMUS 600, and a newly developed payload for the 3000-meter-rated REMUS. The Voyis system includes a 12-megapixel CCD sensor with a 16-bit dynamic range, real-time image correction, and synchronized LED panels for uniform illumination. Robot Operating System (ROS) middleware is used for real-time command and control of the camera system and the automated geotagging of imagery with the vehicle's inertial navigation system (INS) data. The time- and geo-tagging of the imagery facilitates spatial referencing of the data for wide area surveys. Workflows for creating 2D orthomosaics and 3D photogram-metric reconstructions involving advanced post-processing techniques are also presented. The workflow requires capturing sequential images with 60–80 percent overlap and applying Simultaneous Localization and Mapping (SLAM) algorithms to refine navigation data, compensating for drift and ensuring accurate reconstructions. Our results show that high-resolution digital elevation models and orthomosaics are now possible for areas that span hundreds of square meters. Navigation methods, including long baseline (LBL) positioning and dead reckoning, are compared, revealing that LBL yields higher accuracy for wide area imaging surveys.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.