Using edge detection techniques and machine learning classifications for accurate lithological discrimination and structure lineaments extraction: a comparative case study from Gattar area, Northern Eastern Desert of Egypt

Satellite images are used, among other functions, for geologic feature extraction. Scientists have developed advanced, cutting-edge, and reliable data extraction processing procedures, such as machine learning and edge detection methods, as a result of the rapid development in the volume of remote sensing data received from various platforms. The use of the optimal edge detection techniques is crucial for mapping the lineaments because it is thought that hydrothermal alteration may be related to faults and fractures that are pathways for hydrothermal solutions. Efforts were made in this direction by evaluation and comparison of different optimal edge detector algorithms using the first principal component analyses of ASTER imagery on Gattar area in the Northern Eastern Desert of Egypt. The lineaments identified in the Gattar batholith show mainly the NNE–SSW and NE–SW and nearly E–W strikes as major directions and the NW–SE strike as a minor direction where the dominant structures controlling the mineralization are NNE–SSW and NE–SW trending faults. The (LoG) method was found the most accurate and well-matched with the field investigation in the study area. Machine learning methods are robust in processing spectral and ground truth measurements against noise and uncertainties. Four popular and recently established classification machine learning methods are applied and compared in the area for mapping the lithologic units, and their accuracy was evaluated using a confusion matrix; they are maximum likelihood classifier (MLC), Mahalanobis distance classification (MDC), neural net classification (NNC), and support vector machine (SVM). The results showed that both (MLC) and (SVM) were better at discriminating all the lithological units in the area. As a result, it was established in this study that the most appropriate methods specified for automatic lineament extraction and machine learning successfully map geological structures and lithological units in the study area and can be achieved in different geographical locations.