Traffic Incident Management
The efficient management of road traffic is a critical economic factor for cities and countries. Management of traffic flow and incidents is often a manual process, and the configuration of large-scale automated monitoring systems is liable to be error-prone and resource-intensive due to the rule-based nature of such systems.
We designed a system which automatically learns normal behavioral traffic patterns from various sensor data sources (videos, detectors, floating car data, etc.) and correlates it with various context parameters such as day of week, time of day, and environmental factors. This solution allows for management by exception, rather than constant monitoring, as well as automated deployment and continuous performance improvements thanks to machine learning and user feedback.
By using IoTA’s video-based anomaly detection and time-series anomaly detection engines, several solutions for traffic and traffic incident management were implemented.
A high number of false alarms is a key reason for why automated monitoring and management systems often fail in operational, real-world settings. The machine learning-based solution, based on IoTA, reduces alert rates by 95% compared to traditional rule-based approaches, while maintaining actually relevant alerts.