[gdal-dev] Compute raster from other rasters using a Bayesian Network

[Ari Jolma]

Hi,

Just a brief note that I'm working on an extension to GDAL, which allows 
one to compute a raster from other rasters using a Bayesian Network - 
cell by cell. To see why would one do that, see for example

https://www.researchgate.net/publication/240682865_Bayesian_Network_Integration_with_GIS

Such an extension is very simple, just go through all the cells of the 
rasters, feed the data to the network, propagate the information, and 
read the wished output from the network into the output raster.

I'm using the Perl bindings to GDAL, Hugin software, and a perl bindings 
to Hugin. The extension is also a Perl module and the initial version at 
https://github.com/ajolma/Geo-GDAL-Bayes-Hugin is a working 
proof-of-concept.

Hugin is proprietary software, so it is a bit of a problem - for example 
creating web-based tools using it requires expensive licenses. I first 
tried with dlib C++ library but ran into C++ related problems - I left a 
question at their sourceforge discussion forum. There are other (free) 
BN software but I have not yet tried those. If anyone has experiece for 
example with linking SMILE to GDAL I'd be interested to hear.

Best regards,

Ari