Integration and simultaneous estimation of kinematic parameters in multi-view image matching of dynamic processes

Classical methods of photogrammetric deformation analysis are essentially a two-step process of spatio-temporal image matching (STM) followed by the calculation of deformation parameters. In many close-range applications, further kinematic information is also available. So far, this has been integrated into the STM in implicit form as motion models and model assumptions. However, non-modelled effects can interfere with the matching and possibly lead to incorrect results. The project aims to develop a valid general approach that integrates explicit information about the object’s kinematics (e.g. rotational speeds) into the high-precision STM and simultaneously determines unknown kinematic parameters. It is expected that this will lead to a significant improvement in the quality of photogrammetric deformation analysis. In addition, it will provide kinematic information as basic data in the further analysis of dynamic applications in industry and the solution of more general scientific problems. Furthermore, the approach allows the kinematic and geometric parameters to be separated, which means that deformed and undeformed surfaces can also be separated. This is a significant improvement over previous methods, especially for applications in which the object can only be observed in deformed states (e.g. rotating wind turbines). Building on established static models using multi-view image matching, a complete description of the kinematic object, and the corresponding image spaces, is formulated. An approach is thereby developed that estimates the unknown model parameters (geometry and kinematics) on the basis of the given information (image sequences, kinematic measurements, and stochastic information). The primary scientific novelty of the project thus lies in the formulation of a closed and general model description as well as in the development of a highly accurate STM method. The solution is made complete by developing a statistical optimisation of the approach, including spatio-temporal outlier detection. The method is developed and validated utilising simulations and laboratory tests. Finally, the applicability of the developed method is demonstrated by testing a real application. Wind tunnel tests with models of a wind turbine, for which extensive measurement data and additional information are already available, serve as the application scenario.