Structural Wall Recovery and Calibration-based Indoor Area Estimation From Multi-height 2D LiDAR Occupancy Maps

This study presents a postprocessing pipeline for restoring structural wall boundaries and estimating calibration-based indoor areas from occupancy grid maps generated by 2D light detection and ranging simultaneous localization and mapping. In cluttered indoor environments, furniture-induced occlusion and partial observation often fragment wall boundaries, making reliable boundary reconstruction difficult. To address this problem, occupancy maps acquired at multiple sensor heights are aligned and integrated to improve the structural completeness of the input map. The integrated map is then processed through frequency-domain structural analysis, adaptive thresholding, probabilistic Hough transform-based line extraction, wall-level reconstruction, and local wall extension for closed-boundary formation. Gateway blocking is retained as an auxiliary rule for possible open-boundary cases. Compared with a ROSE²-based baseline under the same integrated-map input conditions, the proposed method produced sharper dominant-direction peaks, slightly improved boundary-to-wall alignment, and reduced the runtime of the closed-boundary formation stage while it maintained comparable area accuracy. In the rectangular test space, the five-layer condition reduced the boundary-to-wall distance from 2.741 ± 0.688 pixels to 1.299 ± 0.800 pixels. After isotropic scale calibration, the estimated area was 20.347 ± 0.072 m², corresponding to a relative error of 0.221 ± 0.268%. The results indicate that multiheight map integration mainly improves boundary-to-wall fidelity and reproducibility across acquisition conditions, while calibration-based area accuracy remains comparable across layer conditions.