Traffic modeling in urban system based on RS improved Four-Step Model

Abstract

The traffic modeling is very critical for traffic control/management and urban planning. Numerous methods are proposed for traffic modeling, in which the four-step model developed well and applied widely to analysis the interaction between land use and transportation system. Four-step model comprised of four parts: trip generation, trip distribution, mode assignment and route assignment. And the first step, determination of the traffic demand, is very important to the accuracy of the prediction. Conventional method defined generation rate (GR) of traffic zones to measure the traffic demand, and they are usually determined according to the land usage of traffic zones by which the zones with the same land usage maintained the same and generation rate. It's reasonable and has been widely used due to the fact the land usage reflect the intensity of human activities. However, in urban system due to the complex activities in land use zones, the zones with the same land use status are possible to suffer different intensities of human activities. As the date limitation, the problem of how to measure the traffic demand in urban system is still unaddressed. Towards this problem, this paper proposed a remote sensing (RS) based four-step model to assist the production of generation rate. In particular, the RS indexes NDVI, MNDWI and NDBI were employed to identify the GR and AR. First, by statistic analysis the NDVI and MNDWI were used to extract the areas where there were no traffic demand, for example the water body and the land covered by plants. Second, the NDBI index was analyzed to reflect the degree of building density by the membership function. The membership of belonging to built-up area was then used to assist the determination of GR and AR. By comparing conventional GR and the GR of proposed method, we found the proposed model reflected the land use intensities and the outcome was finer than that of conventional method. The integration of RS and traditional traffic modeling can improve the accuracy and address the problem of data limitation in traffic modeling which would be meaningful.