DoRIAH - Domain-adaptive Remote sensing Image Analysis with Human-in-the-loop
Analysing remote sensing images (i.e., aerial or satellite images) on a large scale requires the balancing of two major constraints: accuracy of results and the time it takes to process the images. While human analysts usually provide highly accurate results, manual analysis is often not feasible in large scale scenarios due to the sheer amount of image data to be processed. Fully automatic image analysis approaches are widely considered as a solution, but often lack the accuracy needed for the specific problem domain. Therefore, modern image analysis methods need to be combined with human supervision to cover a wide range of applications.
The detection of small-size objects in remote sensing images is a common goal in many different application domains: for instance, detecting bomb craters in aerial images from World War II is essential for carrying out unexploded ordnance (UXO) surveys. Inmodern day satellite images, the detection of vehicles provides a rich information source for traffic monitoring or parking lot analysis.
The interactive process for the analysis of remote sensing images involves two basic steps:
- georeferencing and 3D reconstruction from remote sensing imagery, and
- interactive detection of objects of interest.
DoRIAH implements these two steps as an iterative process with feedback loops designed to introduce human cognitive power into the process. In turn, visual feedback methods allow for the efficient interpretation of intermediate results.
DoRIAH is a joint research project conducted by three TU Wien research groups:
- the Computer Vision Lab,
- the Centre for Visual Analytics Science and Technology, and
- the Photogrammetry Research Group.
The company Luftbilddatenbank Dr. Carls GmbH is interested in the efficient analysis of historical aerial images and acts as a business partner.
The project is currently in its initial phase. The first, recently completed, user workshop was aimed at analysing future user requirements. This workshop included external user groups from Austria, Germany and the Netherlands and enabled a comprehensive end user requirements analysis.
Despite the heterogeneous composition of the user groups, concrete applications have been identified in which DoRIAH can increase the efficiency and quality of the analysis results, for example when creating historical elevation models or analysing changes between two or more successive aerial images.
Computer Vision Lab, Institute of Visual Computing & Human-Centered Technology, TU Wien
Other consortium partners
- Research Group Photogrammetry, Department of Geodesy and Geoinformation, TU Wien
- Centre for Visual Analytics Science and Technology, Institute of Visual Computing & Human-Centered Technology, TU Wien
- Luftbilddatenbank Dr. Carls GmbH
Faculty of Informatics
Institute of Visual Computing & Human-Centered Technology
Computer Vision Lab
Tel.: +43 (1) 58801-193176