Vehicle monitoring for traffic surveillance and performance using multi-sensor data fusion

As traffic surveillance technology continues to advance, more complete and intelligent traffic information are available by processing detector data. The proposed dissertation aims to investigate thoroughly the use of recent detector technology in order to obtain useful freeway performance measurements by integrating multi sensor data fusion with a vehicle monitoring algorithm. Vehicle monitoring refers to the identification of the same vehicles at different locations. In this dissertation, two different state-of-the-art traffic detectors are introduced and the subsequent datasets are fused in order to obtain a more robust and effective traffic dataset for vehicle monitoring. Algorithm development for data fusion and real-time vehicle monitoring for traffic surveillance and performance, TRASURF (TRAffic SURveillance and performance) is developed, described, and investigated. Examinations of feature vector extraction from each advanced traffic sensor, fusion across multiple technologies, and analysis of sensor performance are major tasks prior to the development of the vehicle monitoring algorithm---TRASURF. A real dataset was used for single freeway section vehicle monitoring algorithm development and evaluation. Based on extensive field dataset analysis, PARAMICS (PARAllel MICroscopic Simulation), a microscopic traffic simulation model, was used for simulated fused data generation. Based on simulation dataset, multi-section vehicle monitoring algorithm, TRASURF, was tested and evaluated. The proposed simulation framework could be of great value for both testing and performance comparison of traffic surveillance algorithms. Developed vehicle monitoring system, TRASURF, will reconstruct individual vehicle trajectories, which will contribute to effective and efficient traffic surveillance. Furthermore, this enables the derivation of a variety of useful traffic information including network-wide traffic information such as path travel times and origin destination matrices. In addition to the multi-section TRASURF development, this dissertation also investigates and describes various applications of new detector data. Moreover, investigations on various applications of advanced detectors in transportation fields, and especially in single loop configuration, are also presented. The benefit of this approach can be explained by the fact that most of the freeway loop configurations in California, as well as in many other locations, adopt the single loop concept. Unavailable directly from conventional loop detectors, accurate traffic data extraction based on advanced loop detectors will make a vital contribution in many traffic operation and management fields.