Point cloud segmentation and construction verification for large-span modular steel structures

The structure modularization leads to significant demand for construction accuracy. To achieve automated construction monitoring, 3D point clouds obtained from scanning have been applied. This paper proposes an automated construction verification method that integrates prior information. The method achieves rough point cloud segmentation, obtains real axial by template matching, and further calculates the node deviation. Compared with other methods, the proposed method can be applied to complex trusses with multiple types of sections without requiring fine data filtering. The algorithm creatively builds the template library and applies template matching to adjust the posture of member point cloud, mitigating the impact of missing perspectives during scanning. To validate this method, a case study involving a hundred-meter truss was carried out. The data experiment has shown that this method exhibits considerable tolerance to occlusion, noise, and data errors. This method achieves automation in construction monitoring and provides foundation for further research. •A method for automated construction verification was proposed.•The prior information is appropriately integrated into the point cloud processing.•Template library is created and applied in template matching to adjust the posture of PC.•The method exhibits strong adaptability in large scenes and complex environments.•The method is not reliant on point cloud filtering and is less affected by data errors.