Extraction of lattice openings of buildings from terrestrial LiDAR data

We propose a method to effectively extract lattice openings (windows and doors) of traditional Japanese buildings from terrestrial light detection and ranging data. First, the points on a wall are extracted and edges that indicate non-flat surfaces are selected by examining shifts that occur when the data are smoothed. Next, wall direction is determined. Following this, the edges are grouped along wall direction according to the distances between points. Finally, lattice openings are selected from among the edge groups. The thresholds required were set by referring to samples; the proposed method was then applied to a Japanese traditional architecture preservation district. The validation results showed that the accuracy to finally extract lattice openings was 91% (F-measure), which is acceptable for disaster simulation. Therefore, the proposed method was found to be an effective approach for accurately extracting lattice openings from point clouds.