Hybrid Optical and Acoustic Sea Floor Mapping, and Some Other Fun Projects
University of Rhode Island
Graduate School of Oceanography
This talk will present a method for creating hybrid optical and acoustic sea floor reconstructions at centimeter scale grid resolutions with robotic vehicles. Multibeam sonar and stereo vision are two common sensing modalities with complementary strengths that are well suited for data fusion. We have recently developed an automated two stage pipeline to create such maps. The steps can be broken down as navigation refinement and map construction. During navigation refinement a graph-based optimization algorithm is used to align 3D point clouds created with both the multibeam sonar and stereo cameras. In the mapping step, grid cells in the map are selectively populated by choosing data points from each sensor in an automated manner. The selection process is designed to pick points that preserve the best characteristics of each sensor and honor some specific map quality criteria to reduce outliers and ghosting. The final hybrid map retains the strengths of both sensors and shows improvement over the single modality maps and a naively assembled multi-modal map where all the data points are included and averaged. Results will be presented from marine geological and archaeological applications using a 1350 kHz BlueView multibeam sonar and 1.3 megapixel digital still cameras.
The other fun projects portion of the talk will present updates on a few other activities in the lab. These include detecting sea floor venting using a structured light laser imaging, real-time visual odometry for a drifting Lagrangian camera and test results from new vehicle for high resolution water column profiling.
Chris Roman is an Associate Professor at the University of Rhode Island Graduate School of Oceanography specializing in underwater imaging and mapping. He leads a research group that works with multibeam sonars, cameras and novel vehicle platforms to document the sea floor. The goal of this work is to develop new techniques to obtain centimeter resolution maps over large areas. The data products Chris and his students produce are used for marine geology, biology and archaeology.
Chris graduated from Virginia Tech with a BS degree in mechanical engineering in 1997. He then completed a MS degree at the University of California San Diego. He obtained a PhD in ocean engineering from the Massachusetts Institute of Technology and Woods Hole Oceanographic Institution Joint Program.