Geological Information Extraction from Satellite Imagery Using Machine Learning

Satellite imagery has become widely available for detecting patterns by applying different deep learning and machine learning techniques to it. This paper will cover water monitoring, land cover and tree cover visualization. The main aim is to establish the effectiveness of ML and the insights provided by it to promote sound governance, better decision and policy making. Python libraries have been used extensively in Earth Observation applications and resources such as land-use patterns, crop-yield patterns, disaster control, and others. Majority of the data has been made available by Copernicus, World Bank's open data and Landsat programs. Various libraries that have been used are shapely, geopandas, rasterio, fiona, numpy, eo-learn, gdal, and s2cloudless. Detailed use of each of these libraries has been discussed in this paper. The process for analysis is as follows - defining region of interest, downloading satellite imagery, setting appropriate path to provide the algorithm (on jupyter notebook) access to the data, cloud masking, extract various features that are required according to objectives, fetch and save results. We have talked about satellite data, but another important aspect is ground- truth data to reference the results obtained from machine learning and deep learning algorithms. To increase the chances of successful prediction a series of steps are executed before drawing out the results including adding reference, sampling, merging features, postprocessing, and detection of changes. The performance majorly depends upon the distribution of data. It is usually a supervised technique. Recent developments have also provided us an option to fully automate the process, manually handle it or create a hybrid model.