|
||||||
08/2020: Cloud-based Geospatial Processing Services - Extracting Features from High-resolution Imagery: An AI-based Processing Service
Here we partner with BigDataEarth.com (Sydney, Australia) and use new AI and deep learning to extract features of the built environment (e.g. buildings, vegetation - trees or grass, water and roads) from high-resolution imagery for very large geographic regions. The following figure illustrates an AI-based image analysis process which turns RGB input imagery into a predicted probability map suggesting where building roofs are likely located (red colours represent probabilities close to one). From the prediction map, generalised building outlines can be derived.
The next figure shows extracted generalised building footprints or outlines (in the form of minimum bounding box in this case) for an area in Sydney, Australia.
Processing Services on Offer We offer efficient image processing and feature extraction service. You may provide your own high-resolution digital imagery for AI-based image analysis. As an option, we can help license imagery directly from ~10 international vendors who can supply sub-meter resolution satellite imagery. Typically, inputs are in GeoTIFF format (for raster imagery) and outputs in Shapefile format (for vector data).
Case Studies - Newcastle, Australia The next two figures show integrated web and desktop mapping where high-resolution imagery, extracted building footprints, detailed contours and modelled surface water flow directions are displayed together, for a coastal area in Newcastle City Council, NSW, Australia. In this way important location information and insights can be mapped and communicated more effectively.
Disclaimer
| Privacy Statement
| © Copyright GeoSage
|
||||||