Assimilation of parking space information derived from remote sensing data into a transport demand model

Kurzfassung

Accurate data on parking spaces and their utilization is important for optimizing traffic management today and will become even more essential in light of upcoming ITS technologies and autonomous driving. For many cities, however, no comprehensive, standardized and up-to-date database exists. In this paper, we present a novel processing chain combining state-of-the-art remote sensing methods with geospatial analysis. Deep neural networks are used for vehicle detection and traffic area segmentation to identify all types of parking areas and their occupancy on aerial image sequences of the city of Brunswick in Germany. A discretization method is formulated to estimate parking capacity and a regression analysis is performed to draw conclusion for areas not covered by aerial imagery. By comparing the number of stopped vehicles with simulation results from a transport demand model, light can be shed on parking related traffic.