Efficient Rebar Layout Estimation Enabled by a Custom Detection Network and Enhanced Visual Measurement

Accurate detection of reserved rebar spacing in precast concrete structures is essential for ensuring proper component assembly and structural safety. However, conventional measurement methods are inefficient, error-prone, and involve potential safety risks. To address these issues, this paper investigates high-precision rebar spacing measurement from a single monocular image under wide-area and noncoplanar conditions. Accordingly, a vision-based single-image rebar spacing measurement method is proposed. The proposed method integrates a rebar section detection model with a multichessboard, height-adjusted planar calibration strategy. A task-aware image enhancement module is introduced to improve detection stability in complex construction environments. Multiple chessboards are used to construct a wide-area calibration plane, and a height compensation mechanism is employed to mitigate scale recovery errors caused by the noncoplanarity between the calibration plane and the measurement plane, enabling quantitative rebar spacing measurement from a single image. Compared with an structure from motion + multi-view stereo (SfM + MVS)-based 3D reconstruction measurement method, the proposed approach reduces the average measurement error by 76.8% and yields a more concentrated and stable error distribution. The proposed method provides a basis for further development of automated rebar adjustment devises with integrated visual feedback, enabling closed-loop automation from measurement to correction in construction.