Multiple Sensors Data Integration for Traffic Incident Detection Using the Quadrant Scan.

Area of Science:

• Transportation engineering
• Data science
• Traffic flow theory

Background:

• Non-recurrent congestion, often caused by traffic incidents, significantly impacts travel time reliability and causes numerous socioeconomic issues.
• Current sensor technologies face limitations in accurately and promptly detecting traffic incidents, hindering effective congestion management.

Purpose of the Study:

• To develop and evaluate a novel algorithm for early and accurate detection of traffic incidents using time series traffic volume data.
• To assess the efficacy of the Quadrant Scan technique in distinguishing incident-induced congestion from recurrent congestion patterns.

Main Methods:

• Utilized a recurrence-based technique, the Quadrant Scan, to analyze time series traffic volume data collected from multiple sensors on an urban highway.
• Integrated data from multiple sensors in each direction to enhance detection capabilities compared to individual sensor analysis.

Main Results:

• The Quadrant Scan algorithm demonstrated superior performance in detecting traffic incidents when integrating data from multiple sensors.
• The method successfully differentiated non-recurrent congestion caused by incidents from recurrent traffic congestion.
• The algorithm achieved real-time traffic incident detection with a minimal delay.

Conclusions:

• The Quadrant Scan is a promising algorithm for real-time traffic incident detection, offering improved accuracy and reduced detection times.
• The technique's ability to distinguish between incident-related and recurrent congestion provides valuable insights for traffic management.
• The Quadrant Scan algorithm has the potential for extension to detect other types of non-recurrent congestion.