Science and Technology

Data quantities are rapidly increasing in many industry sectors due to the development of new sensors, mobile and cloud technologies, advancements in IoT and AI, and growth of social and entertainment media. Many applications (e.g. in finance, healthcare, government data) have strict information security requirements for simultaneous access, record updating and validation in an immutable manner, which can be achieved with distributed ledger technology (DLT). Sixty one currently available blockchain (BC) systems and their components in context of the DLT for big data storage, provenance tracking, replication, and sharing. The key BC system components and architecture for major information security concerns were elaborated. See please pre-recorded presentation:

https://www.youtube.com/watch?v=n7f_vJKKEV8&

Conference paper available: https://prip.by/2021/assets/files/papers/2124092021PRIP_proceedings_A4-3.pdf

Deep learning techniques based on convolutional neural networks (CNNs) (YOLOv4 and Mask R-CNN) were applied for ship detection in VHR electro-optical/infrared satellite images. They achieved average precision of 94-95% with 3% of false positives without the need of accurate land and cloud masking. Mask R-CNN also allows accurate determining ship size parameters. Watch please:

https://www.youtube.com/watch?v=hAPWDaJoHqU

News on this work in Frontier Energy

the world’s first publication on sectors working in harsh, ice environments (P.16-17 in Frontier Energy magazine).

Stereopol data fusion. The application of polarimetric data in the stereo-radargrammetric process is studied using Radarsat-2 Fine Quad data for digital surface model (DSM) generation. The polarimetric decompositions help extract additional information about the dominant scattering mechanism in the terrain cover for DSM generation. The statistical measures which are potential indicators of DSM accuracy were found.

See please details in the paper

• There is the possibility of theoretical optimization of image (signal) restoration by predicting the required number of images to be processed, the point spread function (PSF), and etc.

  • The experimental results show the recently developed image restoration method enables super-resolution.

This Super-Resolution technique improves fusion results for Remote Sensing.

The 3D reconstruction from several video streams of water flow was used for water analogue modelling of metal casting processes. The low-cost image acquisition equipment can be used. The non-linear recursive method which takes into account a priori information was applied for 3D reconstruction. The experiments with different mould geometries show the high efficiency of Water Modelling.

An analytical expression for the MTF of FPA with arbitrary pixel's geometry (Figure) is found. The MTF of FPA is equal to a half of squared Fourier-transform from function response of a single pixel when there are no frequencies beyond the Nyquist frequency in the object's spatial spectrum. In general case, when there are such frequencies in spectrum, they were determined. The concept of spatially averaged MTF has been introduced.