High Resolution Remote Sensing to Support Transport Infrastructure Monitoring

Remote Sensing can contribute in many ways to the topic of Transport. In this contribution we will focus on several aspects on how high and very high resolution data from satellites and aircraft can generate meaningful information for the monitoring and mapping of traffic related infrastructure objects. In principle there are two main fields: monitoring static objects like roads, lane markings so called road furniture, and other 3D objects and on the other hand tracking of moving objects like cars, trucks, cyclists and pedestrians is of interest. From a spatial resolution point of view airborne data are more suitable for these tasks but it could be shown that also high resolution satellite data can already partly be used for these monitoring and mapping tasks. Through recent developments in machine learning (especially deep learning) the accuracy and completeness of detecting and mapping traffic related objects in image data has increased significantly. Several methods and results are shown for the classification and object extraction of cars, other vehicles, ships, road areas, lane markings, bridges and further infrastructure. Since high definition (HD-)maps are of specific importance for autonomously driving cars a further focus is on generating these HD-maps especially by improving the absolute geo-location accuracy of optical data sets through ground control from precisely geocoded TerraSAR-X data. The potential of using VHR satellite data like from WorldView-x satellites is additionally explored. The task of tracking objects like vehicles and pedestrians in high temporal resolution image time series is shown using airborne image data acquired from airplanes and helicopters. It is demonstrated, how accurate trajectories of single moving objects can be derived and how this can be used for validation of e.g. GPS tracks of cars and also for the in-situ measurements from vehicle sensors. The latter is especially important to be able to independently verify the “car vision” results of autonomously operating cars.