Machine Learning Algorithms for Satellite Image Classification Using Google Earth Engine and Landsat Satellite Data: Morocco Case Study

Earth observation data have proven to be a valuable resource of quantitative information that is more consistent in time and space than traditional land-based surveys. Remote sensing plays vital role collecting many aspects life, whether scientific, economic, or political. Land cover very important supporting urban planning decision making provides opportunities for mapping monitoring areas. Multiple sources exist, including satellite different resolutions ranging from high medium resolution, as well aerial drone image acquisitions. Today, accurate land demand the use imagery remote techniques development becoming common study conducted by researchers find practical solutions problems affecting our planet. The recovery, management, analysis these large amounts pose considerable challenges. classification images popular complex topic. In studies over last decade, been frequently studying only those three machine learning algorithms RF, CART SVM applied on cities countries except Morocco which poses great lack Morocco. To solve challenges, six were compared each other based several evaluation metrics then, avoid download storage space, we used Google Engine, geospatial processing platform operates cloud. It free access substantial computations monitor, visualize, analyze environmental features at petabyte scale. this paper, Landsat 8 perform Morocco, applying algorithms, subfield artificial intelligence. This paper proposes an experimental supervised namely Support Vector Machine (SVM), Random Forest (RF), Classification Regression Trees (CART), Minimum Distance (MD), Decision Tree (DT) Gradient (GTB), order classify water areas, built-up cultivated sandy barren areas forest Moroccan territory deduce end best performing classifier has higher accuracy. results are displayed using set accuracy indicators, overall (OA), Kappa, user (UA) producer (PA). We obtained 0.93 minimum distance (MD) algorithm, but worst result 0.74 support vector (SVM) algorithm. improve results, added indices such normalized difference vegetation index (NDVI), accumulation (NDBI), bare soil (BSI) modified (MNDWI). general, addition improves When comparing classifiers before after indices, yields nearly 93% better Therefore, conclude it was among can quickly produce maps, especially hard-to-reach