You can read the aerial photography in a high level language (e.g. MatLab, IDL or R etc) as 2-D matrices. The thermal and radar data should be similarly straightforward to read in any of those languages, so if you can save the GIS line data in some format that can be read in whichever language you chose then it shouldn't be a problem to do all your processing in a single computer programme. Once you've chosen a language, just have a look at what formats it knows how to read.

That is really helpful, thank you so much! I will report back on how I get on, most probably using R.

To my experience, R is really convenient for modelling but has performance issues in processing of larger image files and GIS data.

A very powerful alternative is python with GDAL/OGR-Bindings (see http://pcjericks.github.io/py-gdalogr-cookbook/index.html#) and scikit-image (see http://scikit-image.org/). The former allows for data import/conversion/export (including vector to raster), the latter offers sophisticated image filtering routines.

I agree with Marian Koch. I'm developing my PhD Thesis here in Germany using scikit for pattern recognition in images derived from LiDAR data. Both Python and Scikit are both very good ways to do what you want. Plus, you don't have to create the whole programming, much of the stuff is already written. All you have to do is to apply in your data. In youtube you can find a lot of videos and tutorials explaining the basics of Scikit-leaning.

Thanks for the useful advice, Joao; I will have to get to grips with Scikit and it is good to know there are youtube tutorials to assist. What did we do before the internet?