Overexposure-aware adaptive scanning and confidence fusion reconstruction for highly reflective workpieces

• Multi-angle pre-scanning predicts overexposure distributions for view quality assessment. • Adaptive main-view and compensation-view planning improves reflective-region coverage. • Pixel-level confidence guides point cloud fusion using fringe modulation, overexposure penalty, and gray-level cues. To address the loss of valid fringe information, unstable view observation quality, and insufficient reliability of multi-view fusion caused by local strong reflection in structured-light 3D measurement of highly reflective metal workpieces, an adaptive turntable scanning and pixel-level confidence fusion reconstruction method guided by overexposure distribution prediction is proposed. The method first obtains the binocular overexposure distribution of the workpiece over the complete rotation range through multi-angle pre-scanning. The main views and compensation views are then planned by considering the overall overexposure ratio, local overexposure concentration, and view coverage relationship, enabling effective view selection for full-view reconstruction. Meanwhile, a pixel-level confidence model integrating fringe modulation, overexposure penalty, and gray-level operating-point constraint is constructed, and the resulting binocular joint confidence is used to guide multi-view point cloud fusion. Experimental results show that the radial RMSE of the reconstructed standard sphere is 0.030 mm. For the regular multi-hole, irregular highly reflective, and multiple-reflection workpieces, the true hole ratios are reduced to 0.106 %, 0.442 %, and 0.530 %, respectively, while the average reconstruction time is reduced by 42.813 %. The proposed method improves the completeness of full-view point clouds of highly reflective workpieces and enhances the efficiency of structured-light turntable scanning while maintaining reconstruction accuracy.