Working group: TLS-based multi-sensor systems

In recent years, laser scanning has established itself as one of the standard procedures in many areas of engineering geodesy, examples of which include deformation analysis and 3D object detection. The task of the working group is to further develop the method as an overall process. The current research work therefore covers the entire spectrum from planning, hardware and platforms, interfaces and evaluation procedures through to final validation. In addition, laser scanning has established itself as one of the most common recording methods for multi-sensor systems, which is also a focus of the working group.

This field of research is concerned with the further development of different kinematic multi-sensor systems that operate on land, in the air or on water. It is essential to research new methods for the reliable and precise calibration and synchronisation of the various sensors with each other, which form the basis for integer georeferencing in any environment. In connection with laser scanner-based multi-sensor systems and the resulting mass data, the focus is, among other things, on the development of efficient filter and equalisation approaches that can be individually adapted to the various application aspects. Thanks to unique laboratory equipment with high-precision sensors for referencing tasks, the developed approaches and systems can be directly analysed with regard to their quality.

For individual sensors and for the combination of different sensors for 3D object detection, these can complement and control each other. However, the sensors differ in terms of their properties (e.g. recording method, accuracy). Accordingly, the uncertainty modelling of multi-sensor systems must include the assessment of the individual data in addition to the consideration of the entire process chain (e.g. calibration and synchronisation of the sensors). High-precision sensors are used to generate reference data and to validate the developed uncertainty models.

Multi-sensor systems generate large amounts of data, meaning that the efficiency of the analysis is becoming increasingly important. At the same time, the context of data acquisition results in individual requirements with regard to analysing this mass data. Against this background, the research field is concerned with the development of customised and generalisable (fully) automated evaluation algorithms based on machine learning.