AI-based Digital Documentation as Sensors of Heritage Morphological Value: Recording Plane Forms of Historical Rural Settlements Hierarchically
Keywords: Cultural Heritage Places, Rural Investigation Workflow, Settlement Plane Form, Computer Vision, Building Regularisation, Geospatial Registration
Abstract. Historical Rural Settlements (HRS) are recognised by ICOMOS-IFLA as tangible cultural heritage, where plane forms preserve multi-layered morphological value accrued over successive eras. These plane forms not only reflect past settlement planning and community interactions but also reveal contrasts between vestigial historical cores and more recent modern developments. Although digital documentation has become a mainstream conservation approach, it often overlooks the interpretive dimension that connects raw geometric data with cultural significance. To address this gap and manage the complexity of large-scale image acquisition, annotation, and analysis, this study treats image-based artificial intelligence (AI) as a cultural sensor for HRS morphology. A re-peatable, open-source workflow that integrates high-resolution remote sensing, expert-driven annotation, hierarchical segmentation (Mask R-CNN), and building(polygon) regularisation was proposed. In the Taihang Baxing (THBX) region of China, 778 HRS samples were processed through a two-stage model with several supportive techniques: the first differentiates historical and modern regions, and the second extracts building footprints within each region. Validation on THBX demonstrates reliable differentiation of hierarchical plane forms and efficient generation of a vectorised digital documentation dataset. By embedding geospatial registra-tion and building regularisation tools, the workflow ensures downstream usability for use cases like geospatial statistics, quantitative morphology and evolutionary mechanisms, providing insights for the rural planning consequently. The proposed workflow emphas-ises heritage morphological value and digital documentation as guiding principles, demonstrating that AI as technical agents can perceptually interpret cultural significance. It can serve as a tool for extracting and highlighting the historical value of HRS in the formulation of conservation strategies and development plans, contributing to sustainable and inclusive rural heritage management.