The traffic monitoring team deals with scientific issues in the field of automatic image analysis for various traffic applications and also for real-time applications for disasters and major events.
Typical applications from the traffic sector are the detection of moving and static objects from remote sensing data, such as vehicle detection and vehicle tracking in image sequences, as well as the area-based segmentation of traffic areas, for example to distinguish between footpaths and cycle paths. For this purpose, both the latest machine learning methods are developed and benchmark data sets are created, which are made available to the public for training and validating own methods.
In the traffic monitoring team, complete systems and process chains including hardware and software are developed. This enables seamless image processing from the image acquisition to the transfer of the evaluation results. The data is usually recorded during flight campaigns with self-developed sensor systems on aircraft, helicopters or UAVs. Especially the airborne optical 3K and 4k camera systems with their high-rate serial images enable a fast recording of large areas as well as an immediate further processing, evaluation and provision of the data. In the past, both current aerial images and information derived from them on traffic or the condition of roads and buildings were able to support authorities and organisations with security tasks (BOS) during disasters and major events.
With the help of the results of the automated process chains, further transport science topics are being worked on, such as the improvement of micro- and macroscopic traffic models (DATAMOST project), the improvement of land transport emission estimation (ELK project), the validation of mobile GNSS receivers (VaGAD project) and the improvement of ego-localisation of road users with the help of aerial image data (KOKOVI project).
Automatic segmentation of traffic areas based on an aerial image