Research On Surface Defect Detection And Evaluation Method Of Optical Curved Surface

Spherical optical components are more and more widely used in civil and military fields,and their quality requirements are also increasing.Surface defects are one of the important indicators for evaluating the quality of optical components,so surface defect detection plays an important role in the manufacturing process of optical components.With the continuous development of artificial intelligence technology,the detection of surface defects of optical components by machine vision has become an increasingly popular technology.some problems.Due to the many types of defects and the different shapes of different defects on spherical optical components,when the machine vision method is used to detect them,the three-dimensional information of the defects is lost and resulting in detection errors.The evaluation of surface defects of spherical optical elements is prone to misjudgment and has poor stability.Therefore,this paper studies the machine vision detection method for surface defects of spherical optical components.Aiming at the problem that the surface defect of the spherical optical element loses the three-dimensional information of the surface defect during the imaging process.The principles of surface defect scattering imaging and spherical three-dimensional reconstruction are studied,and a detection method combining machine vision and three-dimensional reconstruction is proposed.Firstly,according to the imaging characteristics of spherical optical elements,a telecentric imaging model was selected to build an image acquisition platform to obtain high-quality surface defect images.Then,the image of surface defect is preprocessed by image processing algorithm.Finally,based on computer vision image reconstruction technology and back projection technology,the defect image is reconstructed in 3D.Aiming at the problem that the traditional image processing method is prone to misjudgment and poor stability in the evaluation of surface defects of spherical optical elements.A surface defect evaluation method combining deep learning and traditional image processing algorithms is studied.First,identify the defects in the image by edge detection algorithm.Then,the minimum circumscribed rectangle algorithm and the minimum circumscribed circle algorithm are used to obtain the characteristic information of surface defects,and the trained YOLOv3 network model is used to classify the defects.Finally,the evaluation of the surface defects of optical components is completed according to the commonly used testing standard US military standard(MIL-PRE-13830B).The detection platform built in this paper can detect surface defects of spherical optical elements with a diameter of 10 mm and a radius of curvature of 64 mm.Experiments show that the accuracy of this method is increased by 83% compared with the traditional imaging method,and it can accurately evaluate the surface defects of optical components according to the standard,which has research significance and practical value.