Automated image pre-processing
During the last 7 years and several major projects we were working hard on development of an automated satellite image parallel processing system which today stands at a core of innovative technology of our company. This is a modular solution based entirely on standard open source software components (GDAL/OGR, GRASS, SNAP, OTB, etc.) and designed for parallel processing of massive batches of optical and SAR satellite data products in a relatively short time and at very low cost. Efficiency of this solution is a key factor generating our profits. Current TRL of the system is 7. Although it was originally designed to process HR optical imagery, recent consolidation of it's algorithm adapted the system for pre-processing of optical and SAR products of Sentinels 1/2/3 and Supporting Missions products (Landsat 8, SPOT 6/7, etc.). The system can work on GTIFF images, but it is mostly designed for XML-based SAFE and DIMAP product packages of the modern de-facto standard.
Sentinel 1 SAR pre-processing is currently focused only on GRD products and applies standard algorithms of orbit adjustment, noise removal, calibration, terrain flattening, ortho-rectification, despeckling and 8-bit transformation. Pictures on the left show raw Sentinel 1 GRD image in it's sensor projection (top left), a pre-processed GeoTIFF image ready for production (top right), a country-scale SAR data mosaic (mid left) and a time-series (mid right) of 2015 weekly SAR data mosaics of E Baltic.
Sentinel 2 MSI pre-processing transforms standard L1C products into L2A (atmospheric correction) and L3A (cloud-free mosaicking). Two images on the bottom line show L3A products of Lithuania in false color composites: a 10 m resolution image based of NIR/Red (left) and 20 m resolution image based on SWIR/NIR (right).