Quality Inspection of Automated Rebar Sleeve Connections Using Point Cloud Semantic Filtering and Geometry-Prior Segmentation

In reinforced concrete structures, the quality of rebar sleeve connections directly impacts the structure’s safety reserve and durability. However, quality inspection is complicated by the periodic distribution of stirrups, concrete obstruction, and noise interference, presenting challenges for assessing sleeve connection integrity. This paper proposes a training-free, interpretable framework for automated rebar sleeve connection quality inspection, leveraging point cloud semantic filtering and geometric a priori segmentation. The method constructs a polar-cylindrical framework, employing hierarchical semantic filtering to eliminate stirrup layers. Geometric a priori instance segmentation techniques are then applied, integrating θ histograms, Kasa circle fitting, and axial bridging domain constraints to reconstruct each longitudinal rebar. Sleeve detection occurs within the rebar coordinate system via radial profile analysis of length, angular coverage, and stability tests, subsequently stratified into two layers and parameterised. Sleeve projections onto column axes calculate spacing and overlap area percentages. Experiments using 18 BIM-TLS paired datasets demonstrate that this method achieves zero residual error in stirrup detection, with sleeve parameter accuracy reaching 98.9% in TLS data and recall at 57.5%, alongside stable runtime transferability. All TLS datasets meet the quality requirements of rebar sleeve connection spacing ≥35d and percentage of overlap area ≤50%. This framework enhances on-site quality inspection efficiency and consistency, providing a viable pathway for digital verification of rebar sleeve connection quality.