To solve the difficulty of 3D curved surface defect detection, a 3D curved optical surface defect detection method based on micro structured-light detection (MSLD) with large field of view (FOV) is proposed to detect typical defects such as scratches and digs. And a multi-parameter calibration method for MSLD is proposed, which does not have strict requirements on the placement of the camera and the screen and can effectively correct the distortion. Simulations verify the feasibility of the proposed method. 3D detection of different types of weak defects on convex and concave surfaces is also performed. The experimental results show that the reprojection error after calibration is less than sub-pixel, and also show that MSLD can achieve comparable accuracy to white light interferometry (WLI). The difference in measured defect depth between MSLD and WLI is less than 0.5 nm, and the difference in measured defect width is less than 1 μm.