Due to the influence of its optical structure, when using the method of machine vision to detect the surface defects of spherical optical components, it is impossible to image the measured surface on one image plane, and the three-dimensional information of the defects is lost during the detection process. detection error.To solve these problems, a detection method combining machine vision and 3D reconstruction is proposed. First, an image acquisition platform is designed according to the characteristics of spherical optical elements to obtain high-quality defect images. Then, the defect image is preprocessed and identified by the image processing algorithm.Finally, based on the computer vision image reconstruction technology and the spherical center projection technology, the 3D reconstruction is carried out on the image of the defect. Experimental results show that this method improves the accuracy of machine vision detection of spherical optical components surface defects, up to 99%, which is feasible and valuable for research.