DEVELOPMENT OF FREEWAY INCIDENT-DETECTION ALGORITHMS BY USING PATTERN-RECOGNITION TECHNIQUES

Two incident-detection experiments were conducted on the Queen Elizabeth Way Freeway Surveillance and Control System in Ontario. A pattern-recognition approach was applied to improve incident-detection algorithms. By considering the true- and false-incident-alarm identification process as pattern-recognition in nature, the maximum-likelihood decision principle was applied to develop an optimum incident-duration persistence test. The false-alarm rate fell from -.09 to 9.96 percent during a nine-month field test experiment. In the second experiment a two-layer committee-machine structure achieved an 85.7 percent detection rate on 28 samples of historical incident data. /Author/