Mineral exploration modeling by convolutional neural network and continuous genetic algorithm: a case study in Khorasan Razavi, Iran

• Published in: Arabian journal of geosciences Vol. 15; no. 21
• Main Authors: Tahmooresi, Mandana, Babaei, Behnam, Dehghan, Saeed
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.11.2022; Springer Nature B.V
• Subjects: Algorithms; Artificial neural networks; Earth and Environmental Science; Earth science; Earth Sciences; Fault zones; Genetic algorithms; Image classification; Image processing; Labour; Landsat; Machine learning; Methods; Mineral exploration; Minerals; Modelling; Neural networks; Original Paper; Remote sensing; Satellite imagery; Surveying; Continuous genetic algorithm (CGA); Convolutional neural network (CNN); Mineral exploration; Iran; Khorasan Razavi
• ISSN: 1866-7511; 1866-7538
• DOI: 10.1007/s12517-022-10889-7

Mineral exploration modeling to identify promising areas is a crucial step in any mineral exploration program. Traditional methods mainly depend on the experience of specialists. As the images of nature are composed of structures with relatively high similarity, special methods should be used to discover generalizable patterns by extracting and learning the data characteristics. In this paper, convolutional neural network (CNN) algorithms were applied to classify the images relevant to the field of mineral exploration. Input data included the segments of magnetic profiles, ASTER, and Landsat 8 OLI images utilized to determine the effect of the cross-sectional area of magnetic bodies on the magnetic profiles and recognize alterations and fault zones, respectively. In the second part, the continuous genetic algorithm (CGA) was integrated with CNN (CGACNN) to systematically tune CNN hyperparameters. CNNs were trained based on the field survey findings. CNN, CGACNN, and image processing codes were written and implemented in MATLAB. The results show that in comparison with conventional methods, CNNs have more accuracy and can be used to explore minerals intelligently to avoid time-consuming and labor-demanding procedures.