Enhanced data assimilation of 4D LPT with physics informed neural networks

According to recent trend of explosive growth of computation power and accumulated data, demand for the deep learning application in various research fields is increasing. As following this trend, remarkable achievements are presented in the experimental fluid mechanics field. One of the most outstanding research is Physics Informed Neural Networks (PINN) Raissi et al. (2020). Physical knowledge, which has been accumulated by humans, is imposed on the neural networks. PINN was used the automatic differentiation for implementing the governing equations as a physical constraint. By utilizing this concept, physical constraints make neural networks finding physical meaning of phenomena instead of simply fitting to the label data.

Quantifying cloud development from geostationary observations

<p>Clouds and their interaction with short- and longwave radiation represent one of the major uncertainties in our understanding of global climate change. The presence of clouds, particularly of bright low-level water clouds, doubles the Earth’s albedo and they are responsible for half of the solar radiation reflected into space.<br>Contrary to spaceborne, polar-orbiting observations which are of great detail at fixed time we focus on spaceborne time-resolved measurements of the Spinning Enhanced Visible and InfraRed Imager (SEVIRI) aboard Meteosat Second Generation. We present an innovative method to track warm low-level clouds. The method widely used in experimental fluid mechanics and known as particle image velocimetry (PIV) [1, 2] relies on basic pattern matching. The principle of pattern matching is usually referred to as cross-correlation. It tells us something about displacements and enables the reconstruction of cloud trajectories. Thereby, we quantify cloud development and in combination with the CLAAS-2 dataset [3] we characterize temporal changes of cloud properties.</p><p><strong>References</strong><br>[1] Keane, R. D., Adrian, R. J.: Theory of cross-correlation analysis of PIV images. Applied Scientific Research <strong>49</strong>, 191–215 (1992). DOI: 10.1007/BF00384623</p><p>[2] Tropea, C., Alexander, L., Yarin, L., (Eds.), F.: Handbook of experimental fluid mechanics. Springer (2007)</p><p>[3] Benas, N., Finkensieper, S., Stengel, M., van Zadelhoff, G.-J., Hanschmann, T., Hollmann, R., Meirink, J. F.: The MSG-SEVIRI-based cloud property data<br>record CLAAS-2. Earth System Science Data <strong>9</strong>(2), 415–434 (2017). DOI: 10.5194/essd-9-415-2017</p>

Calibration of multiple cameras for large-scale experiments using a freely moving calibration target

Obtaining accurate experimental data from Lagrangian tracking and tomographic velocimetry requires an accurate camera calibration consistent over multiple views. Established calibration procedures are often challenging to implement when the length scale of the measurement volume exceeds that of a typical laboratory experiment. Here, we combine tools developed in computer vision and non-linear camera mappings used in experimental fluid mechanics, to successfully calibrate a four-camera setup that is imaging inside a large tank of dimensions $$\sim 10 \times 25 \times 6 \; \mathrm {m}^3$$∼10×25×6m3. The calibration procedure uses a planar checkerboard that is arbitrarily positioned at unknown locations and orientations. The method can be applied to any number of cameras. The parameters of the calibration yields direct estimates of the positions and orientations of the four cameras as well as the focal lengths of the lenses. These parameters are used to assess the quality of the calibration. The calibration allows us to perform accurate and consistent linear ray-tracing, which we use to triangulate and track fish inside the large tank. An open-source implementation of the calibration in Matlab is available. Graphic abstract