Digital conservation in Tai O Village : point cloud space and architecture conservation legibility

Abstract

This paper presents the results of digital documentation and conservation work conducted in Tai O Village, Hong Kong. The premise of the paper is to explore digital LiDAR and photogrammetry scanning workflow and viability for architectural documentation, analysis, and conservation. Tai O is famous for stilt houses, some of which remain from the 1800’s. Despite their architectural and cultural value, depopulation and policy constraints threaten their future. This situation prompted a research initiative to digitally document and conserve a sample of houses using 3D scanning. The objectives of the project include to digitally conserve the stilt house conditions with 3D scans; to develop and advance techniques for using 3D scans in digital architectural conservation; and to develop and advance techniques for extracting descriptive and analytical information from 3D scans. Just over three hundred stilt houses remain in Tai O, and this project samples twenty selected through snowball sampling. Where possible, researchers sampled stilt houses across the four types remaining in the Village. Researchers assigned house cases a random number and scheduled initial and follow-up meetings to complete data collection in the field. For each case, researchers conducted 1-hour semi-structured interviews with residents before completing a 3D scan of the house interior using a Leica BLK2Go handheld photogrammetry and LiDAR environment scanner. This renders a “point cloud” file of stilt house conditions, which researchers exported to .e57 file types for manipulation in Rhinoceros 3D. The primary contributions presented are workflow for conversion of point clouds to simplified volumetric nonrational b-spline (NURBs) geometry and traditional orthographic architectural drawings; the presentation of a volumetric analysis algorithm based on incremental contouring; and the extraction of meaningful programmatic and physical condition data from resulting architectural drawings. The results find that, for conservation purposes, the legibility of 3D point cloud data gathered using these techniques is adequate for effective, though labor-intensive, translation of orthographic drawings; that the meaning of the volumetric analysis algorithm remains unclear, and that 3D point cloud data allows researchers to extract information required for more traditional typological analysis. Researchers extract cloud data using proprietary software which does not support detailed architectural drafting. File size and compatibility demands necessitate decimation reduction of original point clouds for use in drafting software. This reduces cloud detail and could hypothetically make interpretation into orthographic drawings more difficult. Researchers found, however, that they could produce architectural drawings from decimated point clouds with only occasional field verification. Researchers also drafted simplified volumetric NURBs models of each house and subjected them to a technique termed Volumetric Development Analysis (VDA). This algorithm constructs a volumetric estimation of house geometry using increasing numbers of orthographic contours, which quantifies relative geometric complexity between house cases, and allows comparison between actual volumetric models and estimated volumetric models requiring gradually increasing geometry and computing power. Researchers propose this may provide insight on relative geometric complexity between houses, but results are yet unclear. Lastly, the project yields comparable descriptions of stilt house floor areas and programmatic composition for traditional typological analysis.