A New Traffic Incident Detection Method Under Low-Volume Condition Based On Automatic Vehicle Identification

Most of the widely used Automatic Incident Detection (AID) methods are based on the basic principle of the California Algorithms which detects incidents by comparing the difference of macro traffic parameters between upstream and downstream. However, these AID algorithms cannot perform well under low-volume condition due to many incidents under low-volume don't cause significant changes on traffic conditions. To improve the incident detection under low-volume condition, the paper proposes a new detection method based on Automatic Vehicle Identification (AVI) technique by using high-resolution cameras. First, the paper introduces the basic theory and characteristics of Low-volume Incidents (LVIs) detection by using high-resolution cameras, analyzes the time-space relationship of vehicles passing different detection sections, designs a elimination method for the false plate identification, and proposes a new AID method for LVIs. Second, the paper studies how the main influence factors, including the detection devices density, accuracy rate of AVI and flow level impact the new AID model's performance. Finally, simulation cases are studied to examine the performance of the new AID model. The case study shows that the new AID model can produce satisfied detection results even with very low detection devices density.