Application of Deep Learning in Regional Geological Lithological Mapping Based on Multi-source Geophysical, Geochemical and Remote Sensing Data

Le résumé fourni par la source

Stratigraphic lithology identification and mapping constitute the foundation of geology. Identifying and distinguishing different lithologies is critical for understanding the development and evolution of geological strata. Traditional manual field geological survey is the most accurate method, yet it suffers from low efficiency and high cost. Remote sensing geology is cost-effective and efficient, but performs poorly in covered areas. In this paper, multi-source data including Geophysics, Geochemistry, and Remote Sensing(GGR) are stacked into GeoTIFF files as the sample dataset. A GeoU-Net network model and a predictive segmentation method suitable for regional stratigraphic lithology classification are proposed. The proposed method and system are applied to lithology prediction and segmentation experiments in northern Gansu Province. Compared with manual mapping results, the predicted rock mass distribution is basically consistent, with an accuracy of 91.2%. The model also shows obvious accuracy advantages over other mainstream deep learning-based image segmentation models.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.