Research On Machine Vision System Designed For Optical Surface Imperfection Inspection Based On Raytracing

Optical elements are not only essential components for advanced instruments and engineering research,but also widely used in common electronic and mechanical products.A number of defect inspection items are included in quality control for optical elements,such as stress and bubbles in the material,surface form,surface imperfection and texture.Surface imperfections randomly arise from manufacture,transportation and utilization,which degrade the performance or appearance of the element.As for frontier optics projects such as inertial confinement fusion and extreme ultraviolent lithography,surface imperfections continue inducing damage to element itself and even the whole optical system when lasers transmit through.Imperfections whose sizes are as small as sub-microns should be detected during such applications.The large aperture of optical element to be inspected inherently contradicts with the tiny dimension of surface imperfection.However,visual evaluation by experienced operators is still recommended by relevant standards and applied in industrial production.Visual evaluation method suffers from low efficiency and repeatability.Automatic optical inspection(AOI)based on machine vision is beneficial to objective evaluation of surface imperfection and several illuminating arrangements are developed for different kind of testing samples.Due to the lack of theoretical modeling and simulation,the imaging of imperfection is unstable and the optimization of optical arrangement is unclear.What’s more,currently AOI instruments are limited to test flat surface and spherical surface with large curvature of radius while aspherical surfaces are employed in optical design at an increasing rate.To explore the principle for imperfection imaging and extend the scope of automatic inspection,research related to the virtual modeling and experimental validation of optical surface imperfection inspection system is carried out in this dissertation.General modeling method of machine vision system for surface imperfection inspection is proposed based on raytracing.The machine vision system is composed of a camera,optical surfaces and lights.On the basis of pinhole camera,parameterized camera ray sample model is established which combines the feature of finite aperture camera and telecentric camera in engineering.The equation for optical surface is expressed by vector and the analytical solution to intersections of rays and optical surfaces is derived.The intersection accuracy of analytical and triangle mesh surface model is compared.Once the modeling of components is completed,the rays are emitted from camera,reflected or refracted by optical surfaces and collected on the light surface.Simulated image is obtained by solving image function according to ray paths.Two virtual models of machine vision inspection system are established for different kinds of optical surfaces.The arrangement and inspection flow are optimized according to simulated images and the imaging of imperfection is improved.The reflectance property of different kinds of imperfections is described by bidirectional reflectance distribution function and corresponding illuminating arrangement is discussed.Imaging system model with stripe illumination for reflective surface inspection is established.Image function is solved by Monte-Carlo techniques.A new fusion method is introduced.On the basis of simulated images,the imaging contrast of surface imperfection is improved by adjusting sampling frame number,fusion interval and so on.Dark-field imaging evaluation arrangement for spherical bi-convex,bi-concave,meniscus and plano-convex/concave lenses is virtually modeled to simulate the blind area due to multiple reflections between lens surfaces.According to the accurate prediction of blind area size and position,lights are divided into several groups and images are captured with lights switched in turn.A fusion imaging method is proposed to eliminate blind area to extend the effective inspection area for spherical lens dark-field scattering imaging.These two examples show the flexibility and extendibility of our method.On the basis of optical surface imaging simulation within single sub-aperture,this dissertation further proposes sub-aperture scanning device and method for rotationally symmetric elements.Motion stages and imaging components required for pose adjustment and scanning are introduced.The range of each ring-like aperture is calculated by raytracing on the constraint of microscope depth of field.The dynamic planning solution to the minimum sub-aperture number required for scanning over whole element aperture is illustrated.The stage operation for scanning is introduced and the travel ranges of various kinds of rotationally symmetric surfaces are calculated.An aspheric lens is served as the sample to perform path planning then each subaperture image is simulated in turn.Mapping between sub-aperture images and full aperture projection image are created in accordance with planned scanning path.The flow to stitch subaperture images into a full aperture image is proposed and its accuracy is validated by stitching simulated sub-aperture images.The experimental layout of imaging system with stripe illumination for reflective surface is introduced.Relevant parameters of devices and optical configuration are listed.The designed field of view(FOV)is 1500mm(H)×1100mm(V)and the object resolution reaches 0.3mm.Image sequence is captured for a reflective rough surface.Imperfections,such as dents,fingerprints and scratches,can be observed clearly from fused image.Experimental dark-field scattering imaging device for spherical lens surface inspection is introduction.The experiment is carried out on 4different kinds of small thin lenses with the minimum radius of curvature of less than 20 mm.The distributions of blind area in captured images correspond well with simulated results.Follow the proposed inspection flow and fused lens images efficiently reduce the blind area while remain sensitive for imperfections.Experiment equipment for rotationally symmetric surface inspection is established.Four lenses including a spherical lens,an ellipsoidal lens and two aspheric lenses are served as sample.Within the travel range limit of multiple stages,the maximum scannable diameter of lens reaches 60 mm.Sub-aperture images are both obtained by simulation and captured through experiment according to the planned path.Blind areas are separated processed and eliminated by grouping lights and image fusion.Then sub-aperture images are stitched into a full aperture image.A tiny scratch on stitched image is further measured using a 7.4× microscope.Its width is about 5μm.Mismatches in full aperture image due to various errors are analyzed.Imperfections are well revealed after error correction,which validates the accuracy of proposed stitch method.