Research On Defect Detection Method Of Small Aperture Optical Lens Based On Machine Vision

Small aperture optical lenses are commonly used in various imaging systems such as mobile phones and micro cameras.However,in the actual manufacturing process,how to control the quality of lenses has always been a technical problem that lens manufacturing companies urgently need to solve.Currently,the detection of defects in small aperture optical lenses is generally carried out through microscope observation,but this method has low detection efficiency.Therefore,studying a detection technology for small aperture optical lenses has important engineering practical value.The main work of this article is as follows:(1)A detection method based on diffraction imaging was designed for the detection of small defects in small aperture optical lenses,achieving the amplification of lens defects.CCD cameras were used to collect defect images on the transparent screen.(2)In response to the problem of uneven distribution of image lighting and low contrast between some small defect grayscale values and background grayscale values,directly enhancing the image will produce poor results.Therefore,this article uses Hough transform and ellipse fitting algorithm based on least squares to achieve contour extraction of ROI in the lens area.The comparison results show that the ellipse fitting algorithm based on least squares is superior to Hough transform.Then the contrast restricted histogram equalization algorithm is used to improve the image contrast of ROI region,which can more accurately analyze and process the features in the region of interest.(3)Due to the presence of diffraction fringes in the lens area of the image,it is not possible to directly detect image defect information.In response to this situation,this paper proposes to establish a concentric ellipse model to achieve partition processing of the lens image.Different denoising algorithms are used to process the central bright spot area,and the most suitable algorithm is found based on mean square error and peak signal-to-noise ratio.Then Otsu algorithm based on improved particle swarm optimization is used to detect defects in the image,and the open close operation is used to remove isolated noise and smooth defects,and the region marking algorithm is used to calculate defect parameters;Bilateral filter is used to denoise the diffraction fringe area,then the center line of the diffraction fringe in the area is extracted,and the position information of defects is identified by detecting breakpoints.(4)Aiming at various types of defects such as scratches,point impurities,and bubbles in optical lenses,using parameters such as circularity,rectangularity,area,and circumference as classification criteria,a Support Vector Machine(SVM)classifier was used to extract and classify defect features,achieving high classification accuracy.Quantitative analysis of optical lenses was completed through the minimum external matrix and camera calibration algorithm.