Image Processing and Data Analytics

Most raw satellite imagery needs geometric correction before it lines up with real-world map projections and coordinate systems, and it often needs further work before it can be analysed. Geoimage's image processing and data analytics service does this work for clients, using its in-house processing team in Brisbane.

The team orthorectifies imagery against client-supplied ground control or the satellite's own location data, using the client's Digital Elevation Model or one of Geoimage's in-house DEMs. With surveyed ground control points, Geoimage states a positional accuracy of ±1 pixel. Other steps include pan-sharpening, mosaicking with colour balancing, haze removal, calibration and atmospheric correction for multi-temporal change detection, reprojection to the client's datum, and export in formats for software such as ArcGIS, MapInfo, Imagine, Google Earth, Global Mapper and QGIS.

For analytics, Geoimage classifies multispectral imagery with pixel-based and object-based machine learning, followed by human quality control. The outputs are thematic maps such as land-cover, vegetation, forestry and mangrove-species layers. Published examples use WorldView Legion 8-band, WorldView-2, Pléiades Neo and Sentinel-2 data.

Geoimage has produced state-wide mosaics for the New South Wales Government since 2004. It also produced three whole-of-state mosaic epochs of Victoria for the Digital Twin Victoria program. Other published clients include the Australian Antarctic Division and South Australia's Department for Environment and Water.