Hybrid Visual SLAM for Underwater Vehicle Manipulator Systems

This paper presents a novel visual feature based scene mapping method for underwater vehicle manipulator systems (UVMSs), with specific emphasis on robust in natural seafloor environments. Our uses GPU accelerated SIFT features graph optimization framework to build map. The map scale is constrained by from mounted stereo camera, and we exploit the dynamic positioning capability of system fusing wrist fisheye camera into extend it beyond limited viewpoint cameras. hybrid SLAM evaluated challenging image sequences collected UVMS deep environments Costa Rican continental shelf margin, also evaluate only mode shallow reef survey dataset. Results these datasets demonstrate high accuracy our suitability operating diverse We contribute public use.