Search Results
Multichannel analysis of correlation length of SEVIRI images around ground-based cloud observatories to determine their representativeness
Images of measured radiance in different channels of the geostationary Meteosat-9 SEVIRI instrument are analysed with respect to the representativeness of the observations of eight cloud observatories in Europe (e.g. measurements from cloud radars or microwave radiometers). Cloudy situations are selected to get a time series for every pixel in a 300 km × 300 km area centred around each ground station. Then a cross correlation of each time series to the pixel nearest to the corresponding ground site is calculated. In the end a correlation length is calculated to define the representativeness. It is found that measurements in the visible and near infrared channels, which respond to cloud physical properties, are correlated in an area with a 1 to 4 km radius, while the thermal channels, that correspond to cloud top temperature, are correlated to a distance of about 20 km. This also points to a higher variability of the cloud microphysical properties inside a cloud than of the cloud top temperature. The correlation length even increases for the channels at 6.2, 7.3 and 9.7 μm. They respond to radiation from the upper atmospheric layers emitted by atmospheric gases and higher level clouds, which are more homogeneous than low-level clouds. Additionally, correlations at different distances, corresponding to the grid box sizes of forecast models, were compared. The results suggest the possibility of comparisons between instantaneous cloud observations from ground sites and regional forecast models and ground-based measurements. For larger distances typical for global models the correlations decrease, especially for short-wave measurements and corresponding cloud products. By comparing daily means, the correlation length of each station is increased to about 3 to 10 times the value of instantaneous measurements and also the comparability to models grows.
Graphene transistors for real-time monitoring molecular self-assembly dynamics
Mastering the dynamics of molecular assembly on surfaces enables the engineering of predictable structural motifs to bestow programmable properties upon target substrates. Yet, monitoring self-assembly in real time on technologically relevant interfaces between a substrate and a solution is challenging, due to experimental complexity of disentangling interfacial from bulk phenomena. Here, we show that graphene devices can be used as highly sensitive detectors to read out the dynamics of molecular self-assembly at the solid/liquid interface in-situ. Irradiation of a photochromic molecule is used to trigger the formation of a metastable self-assembled adlayer on graphene and the dynamics of this process are monitored by tracking the current in the device over time. In perspective, the electrical readout in graphene devices is a diagnostic and highly sensitive means to resolve molecular ensemble dynamics occurring down to the nanosecond time scale, thereby providing a practical and powerful tool to investigate molecular self-organization in 2D. © 2020, The Author(s).
Classification and quantification of pore shapes in sandstone reservoir rocks with 3-D X-ray micro-computed tomography
Recent years have seen a growing interest in the characterization of the pore morphologies of reservoir rocks and how the spatial organization of pore traits affects the macro behavior of rock–fluid systems. With the availability of 3-D high-resolution imaging, such as x-ray micro-computed tomography (µ-CT), the detailed quantification of particle shapes has been facilitated by progress in computer science. Here, we show how the shapes of irregular rock particles (pores) can be classified and quantified based on binary 3-D images. The methodology requires the measurement of basic 3-D particle descriptors (length, width, and thickness) and a shape classification that involves the similarity of artificial objects, which is based on main pore network detachments and 3-D sample sizes. Two main pore components were identified from the analyzed volumes: pore networks and residual pore ganglia. A watershed algorithm was applied to preserve the pore morphology after separating the main pore networks, which is essential for the pore shape characterization. The results were validated for three sandstones (S1, S2, and S3) from distinct reservoirs, and most of the pore shapes were found to be plate- and cube-like, ranging from 39.49 to 50.94 % and from 58.80 to 45.18 % when the Feret caliper descriptor was investigated in a 10003 voxel volume. Furthermore, this study generalizes a practical way to correlate specific particle shapes, such as rods, blades, cuboids, plates, and cubes to characterize asymmetric particles of any material type with 3-D image analysis.
Computational tissue staining of non-linear multimodal imaging using supervised and unsupervised deep learning
Hematoxylin and Eosin (H&E) staining is the 'gold-standard' method in histopathology. However, standard H&E staining of high-quality tissue sections requires long sample preparation times including sample embedding, which restricts its application for 'real-time' disease diagnosis. Due to this reason, a label-free alternative technique like non-linear multimodal (NLM) imaging, which is the combination of three non-linear optical modalities including coherent anti-Stokes Raman scattering, two-photon excitation fluorescence and second-harmonic generation, is proposed in this work. To correlate the information of the NLM images with H&E images, this work proposes computational staining of NLM images using deep learning models in a supervised and an unsupervised approach. In the supervised and the unsupervised approach, conditional generative adversarial networks (CGANs) and cycle conditional generative adversarial networks (cycle CGANs) are used, respectively. Both CGAN and cycle CGAN models generate pseudo H&E images, which are quantitatively analyzed based on mean squared error, structure similarity index and color shading similarity index. The mean of the three metrics calculated for the computationally generated H&E images indicate significant performance. Thus, utilizing CGAN and cycle CGAN models for computational staining is beneficial for diagnostic applications without performing a laboratory-based staining procedure. To the author's best knowledge, it is the first time that NLM images are computationally stained to H&E images using GANs in an unsupervised manner.