Supplementary code for Sture et Al. 2019 "Obtaining Hyperspectral Signatures for Seafloor Massive Sulphide Exploration"
This github repository contains the following python scripts (https://github.com/oysstu/sture2019a_minerals). Only python 3.6 or newer is supported. Python must have the following libraries installed.
- h5py (tested: 2.9.0)
- imageio (tested: 2.5.0)
- matplotlib (tested: 3.1.1)
- numpy (tested: 1.17.1)
Optional interaction with Gaussian process models requires the following additional packages.
- tensorflow (tested: 1.14.0)
- GPFlow (tested: 1.5.0)
- pandas (tested: 0.25.0)
- scikit-learn (tested: 0.21.3)
The default paths of the scripts in this repository expects a folder called data with the following subdirectories; masks, model and samples. These folders contain png-files denoting the masks in which reflectance curves are calculated from, a Gaussian process model / calibration data, and UHI data from the respective samples. The paths can be modified in the scripts if necessary.
The following main scripts are available
Downloads a zip archive containing UHI data and pre-computed calibration data. The archive is extracted in the project folder.
A simple viewer in matplotlib for the UHI-data with a slider to change the displayed band.
Create the reflectance plots in figure 6 and 7 in the paper.
By default, this is calculated based on calibration data contained in the sample-files themselves. Optionally, a path to a Gaussian process regression model can be specified (stored as .h5 and .pkl).
Using calibration data from an inclined reference plate, computes a Gaussian process regression model of the measured irradiance over the plate as a function of altitude/height and field of view (viewing angle).