An automatic incident detection (aid) system, using computer based algorithms to identify the traffic disturbances following an incident, is being developed at TRRL. It incorporates two new algorithms, hiocc and patreg. The hiocc algorithm operates by detecting stationary or slow moving vehicles over individual vehicle detectors. It looks for several consecutive seconds of high detector occupancy to detect queues and incidents in high traffic flows. Off-line tests with recorded data show hiocc's detection ability to be very good and a low false alarm rate can be achieved. The patreg algorithm measures the average traffic speed between upstream and downstream detector stations using a pattern recognition technique, detecting incidents by identifying significant speed changes. Preliminary tests suggest that it performs satisfactorily in free-flow conditions up to about 1500 veh/h per lane, and has an acceptably low false alarm rate. By incorporating both hiocc and patreg in an aid system a wide range of traffic conditions can be catered for. A new aid site on the m1 in Bedfordshire will be used to complete the development and testing of the algorithms. (Author/publisher)
Abstract