NSF/USDOT ICSST: Exploring New Traffic Characteristics and Performance Measures Using Feature Extraction and Texture Characterization of Spatiotemporal Traffic Contour Maps

点击翻译按钮获取中文摘要

Information and communication technologies continue to improve transportation data acquisition and performance monitoring capabilities for surface transportation systems. Currently, little information has been extracted from the massive amounts of data being collected in real time by traffic management centers (TMC) nationwide. Despite tremendous advances in transportation data collection technologies, no parallel efforts have been made to improve the existing performance measures by maximizing the utility of information extraction methods from both archived and real-time transportation data. Current measures in the form of point estimates of average travel time and delay are essentially properties derived from first-order statistics and do not adequately reveal certain local and system-wide properties such as smoothness, coarseness, regularity, homogeneity, entropy, and others. Such properties can only be unraveled by advanced procedures and second-order statistics that are capable of characterizing the quality of traffic operations by exploiting spatiotemporal dependencies and textural features of constructed spatiotemporal traffic contour maps, a concept that is similar to texture characterization of digital images. The primary objectives to be achieved in this research are: (1) develop a new class of performance measures by quantifying special features of spatiotemporal contour maps; (2) study the behavior of traffic during transient stages using the new performance measures; (3) establish a procedure for on-line and off-line performance assessment for extended freeway segments; (4) investigate the correlation between the new measures and accident frequency and severity; (5) compare textural characteristics of recurrent and non-recurrent conditions; and (6) develop an implementation module of the new measures to be executed in real time by TMCs. The project examines spatiotemporal freeway traffic contour maps to improve the existing freeway performance measures and to identify precursor conditions for freeway congestion and crashes. Undergraduate students will be strongly encouraged to collaborate with graduate students in the various project research activities, which will introduce them to innovative basic/applied research techniques and strengthen their computational and programming skills. Graduate students will have a leading role in the project and will find opportunities to work on contemporary research topics for their theses and dissertations. The activities and findings of this project will lead to collaborative research opportunities with other disciplines at the intersection areas of transportation, information technology, and computer science/engineering. It is also envisioned that this research study will potentially spark other ITS research activities that would extend benefits to both transportation system users and providers nationwide and within the state of Louisiana.