Streamlining the Evaluation of Potential Marine Debris Targets for Disaster Response

摘要

Hurricanes generate massive amounts of marine debris in coastal waterways, causing environmental problems and hazards to safe navigation. Before debris can be removed, they must be located and identified, a tedious and time-consuming process. Automatic target recognition algorithms speed up this process by searching vast swaths of survey data to locate anomalies that could be debris. However, these targets must still be manually evaluated by analysts. We present a tool that increases analysts' efficiency by automating most of the interaction steps involved in marine debris evaluation. It automatically generates and displays multiple optimal views, which can remove the need for analysts to manually reposition their viewpoint. We also applied these optimal views to a web-based crowdsourcing application, which tested whether the general public could assist in marine debris evaluation. Our results showed that, even though the participants were untrained, their collective decisions were generally in agreement and reliable enough to greatly reduce the number of marine debris targets that must be evaluated by the limited pool of trained analysts. This has the potential to increase analytical capacity in time-critical disaster response situations.