Recognition of Geothermal Surface Manifestations: A Comparison of Machine Learning and Deep Learning

Geothermal surface manifestations (GSMs) are direct clues towards hydrothermal activities of a geothermal system in the subsurface and significant indications for resource exploration. It is essential to recognize various GSMs potential energy However, there lack work fulfill this task using deep learning (DL), which has achieved unprecedented successes computer vision image interpretation. This study aims explore feasibility DL model recognition with photographs. A new dataset was created GSM by preprocessing visual interpretation expert knowledge high-quality check after downloading images from Internet. The consists seven types, i.e., warm spring, hot geyser, fumarole, mud pot, alteration, crater lake, one type none GSM, including 500 different photographs each type. results GoogLeNet were compared those three machine (ML) algorithms, Support Vector Machine (SVM), Decision Tree (DT), K-Nearest Neighbor (KNN), assessment metrics overall accuracy (OA), F1 score (OF), computational time (CT) training testing models via cross-validation. show that retrained transfer advantages accuracies performances over ML classifiers, highest OA, biggest OF, fastest CT both validation test. Correspondingly, selected classifiers perform poorly due their low small long CT. suggests pretrained network be feasible method GSMs. Hopefully, provides reference paradigm help promote further research on application state-of-the-art geothermics domain.