Automatic incident detection (AID) with thermal cameras
Erik P. M. Vermeulen · 2014
Video detection is an established technology for traffic incident monitoring. Through real-time analysis of camera images, a wide variety of incidents can be detected within seconds: stopped vehicles, wrong way drivers, queues, slow moving vehicles, fallen objects ... Video-based incident detection systems only need a short time to detect road events. Fast incident verification can seriously reduce the impact of the incident and prevent secondary accidents. Although video detection has a performance rate of more than 98%, significant issues may arise in some specifically challenging outdoor environments. Due to their exposure to the outside, video cameras can have difficulties coping with varying lighting conditions, reflections of the sun, fog or snow, which may result in false alarms by the detection system. To improve the overall performance of an automatic incident detection (AID) system in challenging conditions or environments, FUIR Intelligent Transportation Systems (formerly Traficon) therefore proposes to combine video cameras with thermal imaging cameras. Thermal cameras see clearly regardless of the glare from sun and headlights, shadows, underpasses, wet streets, snow and many other conditions. This way, it is able to provide 24-hour uninterrupted detection of motorized vehicles and pedestrians. In this paper, a multi-sensor approach towards automatic incident detection is presented, including conventional visual and thermal based cameras. It is demonstrated that this new approach can significantly improve the performance of an AID system, resulting in more reliable detection results and fewer false and missed calls.