USGS COMPUTER VISION PROJECT PROVIDES SUB-MILLIMETER RESOLUTION MAPS OF LAKE-BED BATHYMETRY AND BIOTA

Elevation point cloud of a 14m diameter circular area of Lake Ontario with color display from the pixel where each point was recovered. White areas indicate mussel shell hash, while green areas show Cladophora covered rocks.

Many pressing Great Lakes natural resource management issues are associated with the lake bottom environment, including invasive species (i.e., round goby and quagga mussels), nuisance Cladophora algae, and restoration of spawning habitats for native fish. A general paucity of high-resolution data about bottom type variation has slowed progress toward understanding and managing these issues.

Since 2017, USGS has been working to develop and test a robotic computer vision system that gathers photographs of the lake bottom. Raw photos are then ingested into a data pipeline that converts them into maps of habitat and species. With support from USGS headquarters, the Great Lakes Restoration Initiative, and Environment and Climate Change Canada, two imaging systems have been built and deployed to gather raw data in the lower four lakes, and data processing algorithms are under development in collaboration with Michigan Tech Research Institute. 

Work in the 2018 field season was largely accomplished by SCUBA divers using a custom underwater camera system rated to 40m depth. Data in the 2019 field season are being collected by an Iver3 autonomous underwater vehicle (AUV) with a 200m depth rating.  Early results confirm that the sub-millimeter images gathered by these systems are useful for (1) identifying substrate types and Cladophora, (2) creating dense elevation reconstructions of bottom (results have accomplished > 1 elevation measurement per mm; see figure), and (3) estimating Cladophora volume.  Algorithms are in development to identify and enumerate round goby.  The robotic computer vision system is intended to support monitoring of hard-to-measure Great Lakes Water Quality Agreement sub-indicators, to ground truth other remote sensing techniques (i.e., multi-beam sonar), and to estimate lake-wide round goby biomass to support fisheries management.

Recent Posts

Categories