Date: October 2017
This repository holds python codes for exploratory and research works during my internship on cloud and snow detection for solar atlas purposes.
The recognition of clouds during day-time is one of the most crucial steps for Solar Resources Analysis. An accurate identification of clouds is needed to compute the clear-sky irradiance background and the attenuation superimposed by clouds. This work addresses the empirical knowledge about spectral channels, and then deals with a bunch of methods for the classification of clouds and snow from multispectral satellite images. A test-based algorithm is proposed, as well as a prototype of Machine-Learning approach.
-
Clone the repo
-
Create and activate a virtual python 2 environment.
-
Install requirements. Public dependencies are listed in the
requirements.txtfile. -
Run the code(s)
Disclaimer 1: The goal was to iterate very quickly over research ideas; it was not to produce a future-proof or production-ready library.
Disclaimer 2: this public repo is a trimmed down version of a private repo, which contains both open-source code and proprietary code that cannot be shared.
Besides the modules listed in the requirements.txt file (e.g. scikit-learn, matplotlib, keras, scikit-image, opencv, ...), three private libs (nclib2, himawari8 and general_utils) are required to run the code.
All the files publicly appearing on this repo were written by myself, for exploratory and research purposes.