Mapping Hidden Heritage: Self-supervised Pre-training for Archaeological Stone Wall Mapping in Historic Landscapes Using High-Resolution DEM Derivatives

The study on DINO-CV presents a significant advancement in the automated mapping of dry-stone walls, which are crucial for cultural and environmental preservation in Australia's Budj Bim Cultural Landscape. Traditional mapping methods are often hindered by dense vegetation and the high costs associated with manual efforts. By utilizing high-resolution Digital Elevation Models (DEMs) derived from airborne LiDAR, DINO-CV effectively overcomes these challenges through self-supervised learning. The framework's ability to learn invariant structural representations across multiple views has resulted in a mean Intersection over Union (mIoU) of 68.6%, demonstrating its effectiveness even with limited labeled data. This research not only emphasizes the cultural significance of dry-stone walls but also illustrates the transformative potential of deep learning technologies in archaeological contexts, paving the way for more efficient heritage documentation and preservation efforts.