Statistical analysis of crashes occurring at intersections in malfunction flash

Abstract

This thesis presents an analysis of the characteristics of malfunction flash incidents based on the Georgia Department of Transportation crash reporting database. Malfunction flash is an unintentional state of flash mode in intersection signal hardware. The flash mode is a signal indication of yellow/red or red/red flash. The flash mode can be due to many issues, such as hardware failure, damage, or storms.

Crash reports are completed by police officers at the scene. After processing by the local jurisdiction reports are sent to GDOT for archiving and analysis. GDOT archives the reports in a PDF image format without editable text. This research will develop a procedure to convert the archived PDF reports to text files using optical character recognition (OCR) software. The developed procedure will extract the description paragraph of the incident from the PDF. The extracted descriptions may then be searched for useful information about the incident. The text files will be run through a filter for keywords, such as; "malfunction flash," "red/red flash," "yellow/red flash," and others. Incidents flagged by the keywords will be reexamined to determine if they are malfunction flash incidents.

The 2006 GDOT incident data base will be used for this effort. From an original possible candidate list of 70,000 signalized intersection incidents malfunction incidents will be identified using this method. A statistical analysis will be completed seeking trends and important characteristics of malfunction incidents.