Technology Research On Surface Scratch Defect Detection For Large Aperture Optical Elements

In recent years,high-precision optical systems have become more and more widely used in various aspects of scientific research,military industry and people’s daily lives.As the core component of the optical system,the surface quality of the optical element directly affects the stability of the optical system.The surface quality of optical components is closely related to the size,quantity,density and other factors of defects in the components.Surface defects are mainly scratches,pits,bubbles,openings,impurities and other defects that still exist after polishing and grinding of optical components.Among these defects,scratches are the most obvious and have the greatest impact on the quality of optical components.The presence of scratches will reduce the beam quality in the optical system.Under the irradiation of high-intensity lasers,it is easy to absorb a large amount of light energy and cause damage to the components.The scattered light generated by surface scratches will also be absorbed by other optical components in the system,causing damage to other optical components may even cause the paralysis of the entire optical system.The main reason for the above-mentioned various hazards is that the control of the scratch defects of the optical elements is not in place,so the inspection of the surface defects of the optical elements is of great significance.In this regard,based on a large-aperture optical element defect detection system,this paper proposes a surface scratch defect detection algorithm to improve the accuracy of scratch detection.The main tasks completed include:1.A correction algorithm for uneven illumination was proposed.In the actual optical inspection system,the influence of illumination factors will cause interference to the information extraction of the image.Scratches in the same area can easily cause differences in detection results due to different illumination distributions,and areas with weaker illumination are likely to lose information.The effect of uneven illumination of the image on the later information acquisition,the uneven illumination correction algorithm proposed in this paper uses a polynomial fitting algorithm to fit the extracted background information,and then the original image and the fitted background are differentiated to achieve the uniformity of the image.2.A weak scratch information enhancement algorithm was proposed.The surface defects have shallow depth and narrow width of some scratches.The gray value reflected in the dark field imaging system is similar to the background gray value,and it is difficult to be recognized by visual or conventional machine vision.In response to this problem,the weak scratch information enhancement algorithm proposed in this paper uses frequency filtering and morphological geometric features to separate the weak scratches from the image,and then further stretch the scratch image to expand the contrast of the image,and finally the weak scratch Significantly enhanced trace information.3.An adaptive threshold extraction algorithm was proposed.In the process of image segmentation,it is necessary to set a binarization threshold to separate the background and the defect,so as to accurately locate the scratch defect.The gray value of scratches on high-precision optical elements is often the same as that of the background.A slight deviation of the threshold will lead to failure in the extraction of defect information.The ideal segmentation threshold in different images is often slightly different,so a single segmentation threshold is not suitable for All images.The adaptive threshold extraction algorithm proposed in this paper solves the ideal segmentation threshold based on the gray value distribution of the entire image,the gray value variance and the distribution law,and realizes the adaptive extraction of the ideal segmentation threshold for each image4.A straight line scratch extraction algorithm was proposed.The Hough Transform line detection algorithm has strong anti-interference ability against noise and high accuracy for straight line extraction.This paper combines the non-maximum suppression algorithm and the scratch connection algorithm to achieve accurate extraction of the scratch area The function of position information,and can identify and connect broken scratches,and then develop into a set of Hough transform line extraction algorithm based on non-maximum suppression.5.The curve scratch extraction algorithm was proposed.In view of the fact that some of the curve scratches cannot be identified by the Hough transform method,this paper proposes a morphological skeleton extraction method based on the trajectory of the scratch direction to detect the curve scratches,using the expansion of the scratches,skeleton extraction and skeleton branch removal algorithms to achieve scratches The connected domains are connected to realize the identification and positioning of the scratched area.6.The inspection experiment of surface defects of large-caliber fused silica optical components was carried out.Through this experiment,the accurate extraction of the surface scratches of the large-aperture optical element was achieved,which further proved the effectiveness and accuracy of the algorithm proposed in this paper.The series of scratch detection algorithms studied in this paper are of great significance to the study of surface defects of optical components,and the software integrating this set of algorithms has been used in actual defect detection systems.