Traffic Technology International | December 2021
Accurate and fast detection of incidents
Citilog’s Al-based incident management solution is helping to reduce incident response times on New York Bridges from minutes to seconds.
Traffic incidents are the biggest contributor to non-recurring congestion. Administration, issues such as debris on the road and vehicle crashes are responsible for at least 25% of all traffic congestion. Combined with natural bottlenecks at tunnels, bridges and highways, these incidents are every driver’s worst nightmare.
Citilog, a video analytics artificial intelligence company, is focused on these high-impact problems. Rather than relying on manually monitoring numerous video feeds or reports from drivers to identify incidents, Citilog’s Incident Management Al automatically detects a range of events. These highly accurate detections enable traffic operators to act quickly, thus improving response times and relieving traffic build ups.
25% The proportion of traffic congestion caused by unpredictable incidents such as debris or crashes.
1. Al automates incident detection in tunnels and on bridges and highways
Real world, real results
That is certainly the case for the New York State Bridge Authority (NYSBA), which among other US customers has been using Citilog since 2016. Citilog’s Incident Management Al has enabled the same size team of operators, who actively monitor camera feeds across five bridges 24/7, to increase their situational awareness and optimize their operations. Instead of discovering traffic incidents several minutes after they occur, the Citilog system sends an alert within seconds.
They can then dispatch emergency services (in the case of an accident), crews to assist with disabled vehicles or debris cleanup, or inform drivers via variable message signs (VMS) of slowdowns or alternate routes.
Incident Management Al can even save lives. In one incident last year, a single car veered up an embankment near a bridge approach in the early morning and turned over. While there were no other drivers to report the crash, the rollover was detected immediately by the system and reported to operators who called emergency services to the scene while the driver was still trapped in the vehicle.
2. Incidents are detected in seconds, alerting operators in real-time
What’s the secret?
Citilog’s Incident Management Al leverages deep learning to enhance the accuracy of video analytics to better classify road users and differentiate them from other objects. The system is “trained” with
a large dataset of real-world traffic video clips to identify vehicles and pedestrians and set them apart from shadows and reflections, thus reducing false positives and delivering unparalleled accuracy. The solution is compatible with a variety of camera manufacturers, leveraging the latest advances in image stabilization, low light capture, bandwidth, and storage. As a result, agencies can automatically detect stopped and slow vehicles, general congestion, wrong way drivers, and smoke and debris in tunnels, often using their existing cameras.
” CITILOG’S INCIDENT MANAGEMENT AI LEVERAGES DEEP LEARNING TO ENHANCE THE ACCURACY OF VIDEO ANALYTICS TO BETTER CLASSIFY ROAD USERS AND DIFFERENTIATE THEM FROM OTHER OBJECTS “
