• Solutions

          DATA

June 5, 2024

Incident detection in inclement weather

Traffic Technology International | March 2024

Automatic incident detection with deep learning sees through storms, improving response times to collisions and enhancing road safety

Every rainfall, snowstorm, patch of ice, and gust of wind brings an increased risk for drivers and the agencies that manage our roads. In the US, inclement weather contributes to 21% of all road collisions, according to the Federal Highway Administration, and in Canada that figure is nearly 30%, data from the Royal Canadian Mounted Police reveals. These sobering statistics underscore the challenges faced by traffic operators. When skies turn stormy, highways, bridges, and tunnels become particularly treacherous, with higher speeds, limited visibility, and unique structural hazards amplifying the dangers.
Filtering weather with deep learning In tackling these challenges, Citilog’s Automatic Incident Detection (AID) emerges as a valuable tool. AID is an Al-powered system that utilizes deep learning algorithms trained on a massive dataset of real-world incidents, including those with adverse weather conditions. Other Al incident detection systems lack this important distinction, their performance degrading in poor weather just when they are needed most. Without a way to filter out rain and snow, real incidents are missed, and false alarms can overwhelm operators. In high-risk, low-visibility situations, minimizing these false alarms helps maintain confidence in the system. With deep learning, false positives are reduced by a factor of 10. By training the algorithm to recognize weather, AID discards these non-incidents and identifies


21%
The percentage of total road collisions in the US that are weather-related, according to the Federal Highway Administration

  1. Deep learning sees through storms, discarding weather to detect incidents accurately 2. Reduced visibility shouldn’t compromise your ability to accurately detect incidents
    accidents and other road hazards as they happen. This allows it to detect incidents quickly and accurately, such as stopped vehicles, wrong-way drivers, congestion, debris, and pedestrians, even when visibility is impaired.
    Deploying the correct resources AID provides traffic operators with real-time information about incidents through alerts and visual displays. This helps them understand the severity of the situation quickly and make informed decisions about deploying resources and managing traffic flow. These accurate and timely alerts reduce the workload on operators and enable them to focus on genuine problems requiring attention, ensuring effective utilization of resources. Most importantly, early detection of incidents translates to faster response times for emergency services, reducing secondary accidents and potentially saving lives.

    “EARLY DETECTION OF INCIDENTS TRANSLATES TO FASTER RESPONSE TIMES FOR EMERGENCY SERVICES, REDUCING SECONDARY ACCIDENTS AND POTENTIALLY SAVING LIVES”
    Bill Weber, VP, sales and marketing, Sensys Networks/Citilog
  2. Recently, in upstate New York a motorist rounded a turn in snowy conditions and spun out, the vehicle overturning on an embankment. Thanks to AID, emergency responders were notified and on the scene to assist in just three minutes from the time of the accident. In another recent instance, a pedestrian was detected walking alongside a highway in the snow at night. This potentially dangerous scenario was then proactively addressed, with local authorities ensuring the safety of the individual. Ultimately, for road users, faster incident detection and response translates to quicker clearance of hazards, reducing travel delays, improving safety, and potentially saving lives. But traffic professionals don’t need to save AID for a rainy day. AID is an all-weather tool, increasing vigilance on roadways all year round. For agencies, it’s an investment in the safety of both their crews and the public by responding to incidents faster and more effectively.
Search