Addressing the challenge of scale discrepancies between point clouds in dual-projected structured light systems， which hinder their direct fusion， this study presents a novel multi-scale point cloud fusion method. First， the dual-projected structured light system is calibrated to determine the intrinsic and extrinsic parameters of the cameras and projectors， along with the relative external parameters of the two systems. The weighted least squares phase expansion method is applied to extract the absolute phase， and point cloud data are computed using the calibrated parameters. Next， the spatial distances between any two points within the overlapping regions of the two systems are calculated to determine the relative scale factors， achieving scale unification between the point clouds. Coarse alignment is then performed using principal component analysis， followed by the derivation of the global optimal transformation matrix using an iterative nearest neighbor algorithm. The source point cloud is subsequently transformed to align with the target point cloud based on the calculated transformation matrix. Finally， the Euclidean distance is employed to evaluate the fused neighboring points， and outlier points are removed using a predefined threshold， ensuring accurate multi-scale point cloud fusion. Experimental results demonstrate that the standard deviation of the plane fitting error decreases by approximately 19.56% after fusion， validating the method’s effectiveness in fusing point clouds and reconstructing object surfaces at varying scales. In conclusion， this research provides a robust solution for addressing scale-induced errors in dual-projected structured light systems and successfully mitigates missing data caused by occlusion.