Abstract:Geological mapping has been used in several applications, including resource assessment, hazard prediction, and land-use planning. However, a substantial gap remains in the current state of research regarding methods for acquiring mapping object labels, recognizing mapping objects, and their practical applications in geological mapping. To enhance the accuracy, research depth, and intelligence of geological mapping, while simultaneously reducing fieldwork demands and mitigating risks for geologists operating in challenging terrains, this study introduces an innovative framework based on deep learning to generate 1:50,000 scale regional geological maps. We utilized a deep neural network architecture with multiple shared feature fusion layers focused on geological routes. Compared with conventional methods and random forest-based lithological mapping approaches, this method excelled in harnessing the synergistic potential of multimodal data sources and extracted high-dimensional spatial features from cross-modal data, yielding a method particularly well-suited for geological mapping characterized by extensive data, fine-grained lithological classification within mapping units, limited samples, and data imbalance challenges. This method enabled the recognition of a diverse range of geological mapping objects and maintained an average recall rate of >90% for these classifications. Generating labels based on geological routes enhanced the practical applicability of this framework. Through rigorous experimentation across multiple 1:50,000 and 1:25,000 geological mapping scenarios, the generated artificial intelligence-driven geological maps closely approximated the level of authentic geological maps, with some mapping objects exhibiting superior distribution and precision compared with manually created maps. This approach is poised to transform traditional geological mapping methodologies and processes while advancing the intelligent application of geological surveys in mineral resource estimation, environmental assessments, and coastal zone mapping.

A feasibility study of applying generative deep learning models for map labeling

Learn, Imagine and Create: Text-to-Image Generation from Prior Knowledge.

Deep learning-based geological map generation using geological routes

Map Generation from Large Scale Incomplete and Inaccurate Data Labels

AutoGeoLabel: Automated Label Generation for Geospatial Machine Learning

A roadmap for generative mapping: unlocking the power of generative AI for map-making

Generate Lanscape Paintings From Text to Landscape Painting Generation and Redrawing Using Deep Learning Model

MapGen-GAN: A Fast Translator for Remote Sensing Image to Map Via Unsupervised Adversarial Learning

Semi-MapGen: Translation of Remote Sensing Image Into Map via Semisupervised Adversarial Learning

Keeping Humans in the Loop: Human-Centered Automated Annotation with Generative AI

Conditional Generative Adversarial Networks for Data Augmentation and Adaptation in Remotely Sensed Imagery

Reinforced Labels: Multi-Agent Deep Reinforcement Learning for Point-Feature Label Placement

Analyzing the Training Processes of Deep Generative Models.

GeoBiked: A Dataset with Geometric Features and Automated Labeling Techniques to Enable Deep Generative Models in Engineering Design

Probabilistic road classification in historical maps using synthetic data and deep learning

Understanding urban landuse from the above and ground perspectives: A deep learning, multimodal solution

A Step towards Automated and Generalizable Tactile Map Generation using Generative Adversarial Networks

Learning to Annotate Part Segmentation with Gradient Matching

Exploring Semi-Supervised Learning for Online Mapping

LABELMAKER: Automatic Semantic Label Generation from RGB-D Trajectories

Point-feature label placement with maximum entropy principle