Manual inspection is still relied upon for rebar counting, spacing measurement, and geometric reconstruction in regular double-layer rebar mesh scenarios. The efficiency is low. The results are easily affected by human error. An automatic rebar detection and digital twin reconstruction method based on 3D laser-scanned point clouds is proposed. First, CSF and SOR filtering were jointly employed to preprocess the raw point cloud and reduce the influence of ground points and outlier noise. Then, global PCA alignment and local PCA feature extraction were used to identify rebar candidate points. The point cloud was further divided into vertical-component candidate points, in-plane rebar candidate points, and other points. On this basis, DBSCAN clustering was combined with similarity-based merging strategies at both the cluster level and the cylindrical-segment level, using directional consistency, spatial proximity, and axial collinearity constraints, to achieve rebar instance segmentation and two-stage fragmented-cluster recovery. The recovered clusters were then further classified into transverse and longitudinal rebars according to their principal directions. Finally, RANSAC-based cylinder fitting is applied to recover the geometric parameters of rebars, construct the digital twin model, and perform rebar spacing measurement and analysis. The results show that, under the current controlled experimental condition of a regular double-layer rebar mesh, the proposed method achieves rebar object identification, geometric reconstruction, and parameter extraction. The identified transverse spacing, longitudinal spacing, and interlayer spacing are all generally close to the design values. This indicates that the method is feasible for this type of scenario. The findings provide a reference for the digital inspection and parameter evaluation of double-layer rebar meshes.