Safety First: LSU Receives $773K Grant for Traffic Study
November 11, 2019
BATON ROUGE, LA – In 2018, there were 163,706 automobile accidents in Louisiana, with 4,086 of those involving a commercial motor vehicle. Hoping to see these numbers decrease, LSU researchers are now working with the Louisiana State Police to predict and prevent crashes before they happen.
The Federal Motor Carrier Safety Administration recently awarded a grant in the amount of $773,504 to LSU’s Center for Analytics and Research and Transportation Safety (CARTS) to study the causes of CMV crashes in an effort to improve safety on Louisiana’s roadways. CMVs include 18-wheelers or any vehicle used to transport goods or passengers for profit.
Jeff Dickey, geographic information systems (GIS) manager for LSU’s Highway Safety Research Group, came up with the idea to use traffic camera footage as a live tool and began working with CARTS, which is responsible for collecting, maintaining, integrating, analyzing, and distributing crash-related data captured from law enforcement and other agencies throughout Louisiana.
“There are a lot of traffic cameras and data out there but no one to collect it,” Dickey said. “The problem is that it’s huge to store that much video data. The idea is to reduce a lot of video data to a series of images or events that can be captured and stored in a database.”
Dickey said the cameras will capture events, such as cars following too close, not maintaining a lane (swerving), hard braking, or someone cutting off a truck or other vehicle.
“We don’t know how many of those incidents are really happening on the road,” he said. “We only find out about the bad ones when there’s an accident. If we can document unsafe driving behavior, we can actually tally the number of events and keep a running total.”
Dickey and LSU Computer Science Associate Professor Supratik Mukhopadhyay want to create a dashboard to spot the high-traffic-volume locations where an accident is likely to occur. The police could watch this dashboard, or the dashboard could be programmed to send an email or alert when there’s an abnormal event.
“We want the cops to know the second a crash happens,” Mukhopadhyay said. “Using artificial intelligence-driven analytics, we can analyze the streaming video to look for abnormal events, such as a car going the wrong way or traffic piling up. We can predict traffic patterns that can lead to a crash. If we can see what leads to a crash, the machine learns from it and says there’s the possibility of a crash here. Then it will inform the police that they should do something in that area.”
“There are a lot of people switching lanes all of a sudden, so maybe there’s something in the road,” Dickey said. “The police could deploy resources to make a presence. Even if they’re just there, people are going to slow down.”
By identifying driver behavior, Dickey and Mukhopadhyay can see if that behavior is associated with interstate crashes. By working with CARTS, who has crash data going back to 2005, it’s already known what parts of the interstate have a higher crash rate, such as the Atchafalaya Basin.
“We have something similar for the Bonnet Carré Spillway on I-10 but they don’t have a high crash rate,” Dickey said. “So, we’re trying to see if there’s something going on that makes a difference. Is it driver behavior or something structural? If we can relate behavior that we see on camera to the types of crashes we’re getting on the road, we can possibly say it’s driver behavior.”
Helmut Schneider, executive director of statistical analysis for LSU’s HSRG and CARTS, will serve as the principle investigator on the CARTS project. His recently prepared data shows that over the past five years, there have been 983,052 vehicle crashes in the state, with 24,125 of those involving a CMV. From January to July 2019, there were already 29 fatal CMV crashes.
HSRG hopes that within two years, there will be a system in place for state police to use.
“They are really excited about this, as are we,” Dickey said.
Contact: Libby Haydel