ABSTRACT
Spalling is one of the typical damage types of concrete structures. Concrete spalling occurs with local concavity and volume loss, leading to a certain range of load capacity reduction. Recently, with the rapid development of structural health monitoring and non-destructive inspection technologies, it becomes possible to perform high-resolution measurements and capture rich details of civil structures. By using laser scanners, Lidar scanners, or photogrammetry techniques, dense point clouds of structural faces can be reconstructed. Those dense point cloudscan describe both global and local geometries of structures in detail, which can be further used for damage assessment. However, the researches on identifying and localizing structural damage on concrete structures based on PC processing are still limited. Therefore, in the article, we propose a method to segment concrete spalling in 3-D point clouds of concrete structures. Unsupervised clustering methods are used to segment the point cloud into several classes that represent different faces of the target structure or structural member. Then a plane is fitted to each class of the point cloud using the Least Squares method. Based on the distances from the points to the corresponding fitted plane, a threshold can be determined, and the points that are far off the plane can be identified and localized. Those identified points can be used to present concrete spalling damage. To validate the proposed method, an experiment was conducted on a reinforced concrete beam. The result shows that the volume loss of the concrete beam was successfully identified and localized. Using a distance threshold, the points of damage can be segmented from the point cloud. Therefore, the feasibility of the proposed method was fully demonstrated.
