Point cloud-based crack detection and measurement in masonry buildings subjected to settlement induced by underground construction

Structural damage in masonry buildings subjected to settlement induced by underground construction often manifests in the form of cracking. In current practice, crack detection and monitoring rely on visual inspections and single point measurements. Research efforts to overcome the limitations of these techniques have focussed on image segmentation and anomaly detection tools, which do not provide information on displacements leading to crack development. This paper introduces an alternative non-contact procedure based on the use of point cloud data. The proposed Point cloud crack analyser (Picca) method exploits the subtle geometric features on masonry surfaces, and recasts the problem of crack detection and monitoring as a motion measurement problem. By characterising the motion around cracks as rigid-body movement of matched feature points before—and after—deformation events, Picca detects cracks via the relative displacements between clusters of rigid-body motions. Investigations with synthetic data reveal that the minimum detectable crack width with this method corresponds closely to the average point spacing. Picca is evaluated using data from a recent test campaign on brick masonry buildings, where crack detection and measurements from point clouds demonstrate good agreement with benchmark results. Since it uses geometry information only, Picca is insensitive to colour changes and light variations. It does not require fine tuning as the input parameters for the underlying feature calculation and robust registration algorithms are automatically set on the basis of geometric and statistical interpretations. These aspects highlight the suitability of Picca for future field applications. The research code is shared with this paper to facilitate its use and is available at: https://github.com/yliu17lhr/pc_cr.