Exploring Google Earth Engine Platform for Satellite Image Classification Using Machine Learning Algorithms

Google Earth Engine is a geospatial data processing platform that runs in the cloud. It offers free access to massive amounts of satellite data as well as unlimited computing power to monitor, visualize, and analyze environmental features on petabyte scale. The capability of this platform to support diverse approaches for land use and land cover (LULC) classification using both pixel based and object-oriented methods has been made possible through the provision of a variety of machine learning algorithms. Earth observation data have proven to be a valuable resource of quantitative information more consistent in time and space than traditional ground surveys. They offer numerous opportunities for mapping and monitoring urban areas, as well as a variety of physical, climatic, and socioeconomic data to support urban planning and decision-making. We used Landsat 8 satellite data to perform supervised classification in this paper, and we used three advanced machine learning methods Support Vector Machine (SVM), Random Forest (RF), and Minimum Distance (MD) to classify water areas, built-up areas, cultivated areas, sandy areas, barren areas, and forest areas on Moroccan territory. The classification results are displayed using a set of accuracy indicators that includes overall accuracy (OA) and the Kappa coefficient. We obtained 0.93 as a better accuracy for the MD algorithm, however, the worst accuracy result is 0.74 for the SVM algorithm.