Within the institute’s strategic workflow for deploying optical systems, the research group for real time image processing is engaged in R&D work targeted towards supplying efficient and robust image processing algorithms and their implementation.
The main objectives of the group are the realization of increasingly intelligent, autonomous, real time-capable optical information systems and the creation of technology for the derivation of high quality data products from high resolution data of optical sensors. The group has a special emphasis on the extraction of 3D information from multiple images.
Such derived depth information enables further data processing steps leading up to object reconstruction with subsequent generation of up-to-date spatial environment maps, or DTM (digital terrain model) generation and following automatic three-dimensional visualization of observed areas.
These image processing techniques have special requirements: exact calibration of the optical systems, use of additional information of other sensors (inertial systems, vehicle sensors etc.) and mastery of photogrammetric methods. Therefore the real time image processing group cooperates closely with the relevant department specialists.
Optimized implementation of high-quality stereo matching algorithms for real-time and high-performance applications is an important part of the group’s focus. The use of special hardware components for example FPGA and GPU enables efficient and cost-effective solutions.
The innovative technologies are relevant to a broad spectrum of applications: on the one hand to real-time computer vision applications, e.g. in robotics, automotive or railway sector, in medicine or airport monitoring, and on the other hand to applications requiring high-throughput analysis of huge amounts of data, such as in airborne or spaceborne remote sensing.