Advanced land imager superiority in lithological classification utilizing machine learning algorithms

Abstract Different types of remote sensing data are commonly used as inputs for lithological classification schemes, yet determining the best source each specific application is still unresolved, but critical interpretations. In addition, various classifiers (i.e., artificial neural network (ANN), maximum likelihood (MLC), and support vector machine (SVM)) have proven their variable efficiencies in mapping, which technique preeminent questionable. Consequently, this study aims to test potency Earth observing-1 Advanced Land Imager (ALI) with frequently utilized Sentinel 2 (S2), ASTER, Landsat OLI (L8) allocation using widely accepted ANN, MLC, SVM, a case Um Salatit area, Eastern Desert Egypt. This area has recent geological map that reference selecting training testing samples required learning algorithms (MLAs). The results reveal (1) ALI superiority over most S2, L8; (2) SVM much better than MLC ANN executing lithologic allocation; (3) S2 strongly recommended separating higher numbers classes compared L8, ALI. Model overfitting may negatively impact classifying small targets; (4) we can significantly enhance accuracy, transcend 90% by blending different sensor datasets. Our new approach help further mapping arid regions thus be fruitful mineral exploration programs.