In order to further develop photogrammetric techniques for 3D monitoring of dynamic processes, the target variables (e.g. deformations, crack widths) shall be introduced as unknown directly into the parameter estimation process, thus combining the geometrical model of photogrammetry with deformation models from mechanics. From a stochastic point of view, this has the advantage of higher redundancy and the implicit use of covariance in the measurement process, leading to an increase in accuracy and reliability. This is particularly relevant when photogrammetric measuring techniques are employed close to the limit of their applicability, as often the case in the evaluation of high-speed camera image sequences of highly dynamic processes. The model-based approach aims to compensate for this impairment and to make stereo high-speed camera systems a universally applicable tool in the analysis of high-dynamic processes.