Provide More Detailed Algorithm Descriptions

[theoilie]

It would be beneficial to have more thorough descriptions of what each part of the program does internally, even if it's at a high level. This would especially make sense for the examples portion so that people can know what exactly it is that they're looking at. Maybe we can compile our knowledge here and then format it nicely in the examples README? We can also include non-essential / lower-level information in separate documentation files.

Information we have so far:

Mineral

This algorithm classifies each pixel as a certain kind of mineral. Once it finishes running, you can launch QGIS (download/install it here) and view it by clicking Layer > Add Layer > Add Raster Layer... Then select the .img file (something like _img_class.img). You can expand the layer, double click on one of the minerals, open the Symbology tab, and then change the colors to visualize it however you see fit.

Mining

Mining runs in a similar fashion to mineral. It takes the output of mineral classification (a classified GIS image) as input and subsets pixels based on a configurable list of materials e.g. Schwertmannite, Renyolds_TnlSldgWet, Renyolds_Tnl_Sludge. The output is a new classified, subsetted GIS image representing pixel containing materials which are associated with mining.

Correlation

The environment correlation part runs gdal_rasterize to match hydrography features with the image that the mining part produced. It does this using the National Hydrography Dataset, which contains information about water drainage paths in America (the flow of rivers, lakes, dams, etc...). In the example we currently use (example_environment.py), the code runs gdal_rasterize.py to match the mining image with NHDFlowine.shp. This shows watershed geometries and the directions water flows in and superimpose it onto the mining image. Read this for more info about NHDFlowline. Read this for more information about the NHD in general. There's also a nice visualizer of United States hydrography data here. It includes NHD and NHDPlus, but I'm not sure if this project supports NHDPlus or how the file format has changed for NHDPlus other than making it higher resolution. To use the correlation algorithm in a meaningful way, you have to download a shape file from this link that is in the same area as the spectral data you're using (it will be in the format NHDFlowline.shp).

@capstone-coal/19-usc-capstone-team

[lewismc]

Thanks for this @Lactem