Foundational Methodology for Creating Supplemental, FHWA-specific, Roadside Element Data Set, toward Using Big Data Techniques

摘要

The National Highway Traffic Safety Administration (NHTSA), an agency of the United States Department of Transportation (US DOT), provides crashworthiness data sets serving as a starting point for highway research. Although the information has provided guidance to safety researchers and decision-makers for vehicle damage and occupant injury research, crash environment fields have been inadequate to serve the growing surveillance burden placed upon the Federal Highway Administration (FHWA), another agency of the US DOT. Recent advances in NHTSA data reporting, however, have allowed for supplemental data acquisition, at no added cost. The 2008 Fatality Analysis Reporting System (FARS) release marked the first instance in which the geographical location was reported for fatal crashes and placed in the public domain. This opened a technological linkage with mapping tools to improve understanding of the crash environment. Further, improved photography introduced in the National Motor Vehicle Crash Causation Study and concurrent improvements in image capture for the National Automotive Sampling System Crashworthiness Data System provided better understanding of the crash scene. This paper provides the manual data extraction approach used in identifying candidate crashes and codification of roadside elements. Currently, FHWA is securing access to the Second Strategic Highway Research Program data, representing large stores of video. Leveraging recent FHWA Exploratory Advanced Research findings, in data mining and video capture, and lessons from the National Aeronautics and Space Administration and National Institutes of Health, the current project seeks to extend the manual approach to big data extraction.