Photogrammetry and Image Analysis Department
The Photogrammetry and Image Analysis department focuses on developing operational image processors, deriving geometric (photogrammetry) and semantic (image analysis) information from digital remote sensing images acquired from satellites, aircrafts and UAVs. The growing number of missions (including the Copernicus program and new satellite constellations), along with the rapid developments in machine learning and the increased computational power in processing chains, are the driving forces behind remote sensing technology development, which led the department to outstanding operational and scientific achievements. An important goal is to provide standardized and high quality data to the growing user community. The pre-processing of raw satellite data is therefore a necessary prerequisite, and is generally included in the ground segment of any mission. In the specific, the department is currently developing the ground processing software for the hyperspectral DESIS and EnMAP missions, which includes calibration, systematic and radiometric processing, orthorectification, and atmospheric correction. We also develop software for several other high resolution optical missions which are integrated into ground segments, and license software to commercial partners.
In the department, the methodological and application-oriented scientific developments for automatic geoinformation extraction from optical or multimodal data include the following research fields:
- Photogrammetry: Stereo processing and object detection
- Hyperspectral data analysis
- Atmospheric correction
- Real-time methods for imagery from airborne sensors for traffic monitoring
The photogrammetric work concentrates on direct georeferencing using position and attitude data (physical approach, also calculated in real time on GPUs), including automatic control point search in image databases and the derivation of digital surface (DSM) and terrain models (DTM) from optical stereo data using computer vision methods. By analyzing the DSM and the extracted image features, objects such as buildings can be detected.
The development of new methods for hyperspectral data analysis plays an important role at IMF, especially with regard to the two missions DESIS and EnMAP. In this context, the department makes important contributions to the evaluation and interpretation of the data. The expertise ranges from in situ land/water spectrometry and atmospheric correction to the development of dedicated algorithms such as spectral unmixing, denoising, fusion with multi-spectral data and semantic analysis.
The department is internationally known for its atmospheric correction processors of optical data, and is now pursuing parallel developments. As a result of recent research projects, the software PACO (Python Atmospheric COrrection) can handle a wide range of optical satellite and aircraft data, building on the experience acquired with its predecessor ATCOR. Together with the French CNES, the best-of software MAJA (MACCS ATCOR Joint Atmospheric Correction) was developed within an ESA project especially for Sentinel-2 data.
By developing a hardware and software system for real-time processing of airborne serial image data directly onboard the aircraft, the department has established another unique development throughout Europe. The system has undergone significant further improvements in recent years and enables orthorectification, DSM creation and mosaicking of image data and, above all, the derivation of traffic-relevant parameters such as vehicle detection and speed detection. The system can also derive pedestrian movement patterns and people density in large crowds. Delivering support for autonomous driving using remote sensing data is a completely new field of research at the institute, which is to be strongly expanded.