Automatic Track Identification:An Adaptive Pattern Recognition Algorithm,

摘要

The identification of vehicle tracks is a primary requirement in the use of data from sensors that are emplaced to monitor vehicle movements. The identification process, whether manual or automated, must contend with false-alarm activations, failures to detect, and other data corruptions resulting from sensor unreliability. The detection algorithm presented here automates the previously manual task of identifying vehicle tracks in the data received from unreliable sensors. It is distinguished by its capability to dynamically discriminate among sensors based on their past performances in detecting vehicles and avoiding false alarms. Extensive testing on live and simulated data indicates that the algorithm performs approximately as well as a trained human operator and that it is somewhat superior to another algorithm previously developed for this purpose.