A Low-Cost Underwater Stereo Camera System for Underwater Quantitative Imaging

Abstract Structure-from-motion (SfM) photogrammetry has become an increasingly useful tool in recent years for underwater surveys of seafloor environments. The current method involves divers placing physical ground control points (GCPs) and scale bars on the substrate, swimming in a boustrophedon pattern over the substrate, and taking several overlapping images with high-cost digital cameras. The necessity of GCPs and divers when using a single camera greatly limits the depths in which spatially rectified SfM photogrammetric reconstructions can be created. The introduction of stereo camera systems, when properly calibrated, can remove the need for GCPs and scaling objects, allowing autonomous and remotely operated vehicles to facilitate quantitative image surveys of benthic environments beyond diver depths. Many ecological applications for using stereo camera systems, including SfM and accurate measurement of mobile fauna, make the necessity of a low-cost underwater stereo imaging system crucial for efficient monitoring of marine ecosystems. This study presents the design of an underwater StereoPi, a compact and low-cost stereo imaging system designed using Raspberry Pi components and encased in a stereolithography (SLA) 3D-printed resin housing, as well as an open-source calibration and image processing workflow using Python 3.9 and Agisoft Metashape Professional. Underwater calibration data, including comparative hydrostatic tests, are presented to determine how hydrostatic pressures affect the system’s accuracy of stereo imagery measurements and spatially rectified reconstructions in deep water. The presented system has been field tested in Bermuda using a fiber optic tethered remote system for real-time disparity map viewing through first-person goggles to a depth of 70 m, as well as a diver-held system in Panamá for benthic SfM photogrammetry.