Automatic detection and site-scale vegetation shift analysis for individual heritage buildings along the Great Wall

Great Wall individual heritage buildings (IHBs) are of historical importance but increasingly threatened by vegetation-induced erosion. Effective conservation requires monitoring vegetation dynamics at the site-scale, yet precise spatial data for IHBs is lacking. To address this, we propose a comprehensive framework for site-scale vegetation shift analysis based on automatic IHB detection. First, we develop IHBSegNet, a deep learning segmentation network with three core modules to accurately extract IHB footprints. We then apply an unsupervised clustering algorithm to classify vegetation density from 10 m satellite embedding data. Finally, per-pixel class shifts are quantified and aggregated within each IHB for vegetation shift analysis. Case studies along the Great Wall sections in Fugu, Shenmu, and Yuyang show that IHBSegNet outperforms state-of-the-art networks, achieving an IoU of 73.02%. Analysis of detected IHBs revealed significant vegetation density shifts at 67 sites, with densification observed in Fugu (26 sites) and Yuyang (25 sites).