Mapping of Himalaya Leucogranites Based on ASTER and Sentinel-2A Datasets Using a Hybrid Method of Metric Learning and Random Forest

Abstract

The widely distributed leucogranite belt in the Himalayan orogeny is expected to have excellent potential for developing rare metal mineralization. Finding a way to effectively map the spatial distribution of leucogranites would be a significant contribution to rare metal exploration. Research shows remote sensing technology has long been recognized the significance in geological works, which greatly promoted mineral exploration in a cost-effective manner, especially in the Himalayan orogenic belt with poor natural environment. However, several challenges still exist in relation to the limited spectral band and spatial resolution of remote sensing images, as well as the onerous data processing. In this context, this study sought to resolve these two issues by applying a hybrid approach that comprises image fusion, metric learning, and random forest methods. For the first challenge, multisource and multisensor remote sensing data were integrated to provide more comprehensive spatial texture characteristics and spectral information. To address the second challenge, this study used a hybrid method of metric learning and random forest to promote computing efficiency and classification accuracy. This process is illustrated through a case study of lithological mapping in Cuonadong dome, the northern part of the Himalayan orogeny belt. Seven target lithological units were effectively discriminated with an 85.75% overall accuracy. This provides an important scientific basis for further exploration for rare metal deposits in the Himalayan orogeny belt, and a way of thinking for detecting geological features under harsh natural conditions.