Research On High Signal-to-noise Ratio Inspection Technologies For Large-caliber Optics Damage

Large-caliber optics damage high signal-to-noise ratio inspection system is the important part of the Inertial Confinement Fusion(ICF)system,Which to ensure the proper functioning of the ICF.The main task of the inspection system is to inspect the optics in the final optics module accurately and fast With high signal-to-noise,identify the damage of the optics.And then to evaluate the health of optics.A set of large-caliber optics damage inspection system is designed in this article based on internal total reflection lighting principle,combining With the relevant machine vision technology.What's more in this article,under the condition of internal total reflection lighting light-field characteristics is analyzed deeply,then the imaging principle of the truth damage and imaging reason of the false damage is researched,finally the key technology of deleting the false damage is researched.The main research Work is as follo Ws:Synthetically analyze the principle of optics damage inspection and the lighting system,design a set of large-caliber optics damage high signal-to-noise inspection system,in vie W of the special structure of the final optics module,designing a proper of point light source and the line light source.And then using the Trace Pro soft Ware calculation the optics internal light field under the both light system,get the result that the mean flux of the surface in the optics under point light source is bigger,and the l ?minous flux distribution in the optics under the line light source is more uniform,and through experiment to validate the results.Calculate the scattering of the damage surface to the total internal lighting using the scattering principle of electromagnetic Wave in the random surface.And calculate the imaging of the lens to the scattering light by the angular spectr?m method.Eventually get imaging characteristics of real damage point.Then get the main reason for the false damage images by both analyzed and experiment.Finally research to delete the false damage in the online image.First,use Local Area Signal to Noise Ratio algorithm to detect all the damages in the online image.Then combine With the data Which is collected by Confocal Microscopy to generate the sample data.Finally use ELM and H-ELM algorithm to train the sample data and get the classifier.By co mparing the classification accuracy of the ELM and the H-ELM,it is determined that the H-ELM algorithm With 1500 hidden layer neurons and the parameter 81?10 that is used for the regularized least mean square calculation has the highe st classification accuracy,Which is achieved to 95.1%.