| Final Optics Module (FOM) is one of the most important components of Inertial Confinement Fusion(ICF) system, which involves a range of massive final optics. Due to factors such as laser’s self-focusing, laser induced damage may arise inside or on the surface of the final optics, which are exposured to radiation from high-energy laser. Concerning the exponential growth in the lateral size of the damage site with shot number, potential defects should be detected and tracked in the initial phase of formation. Maintenance is needed before the defect size reaches a certain threshold; otherwise, the damage would be irreversible. The present research concentrates on final optics online inspection technology.Considering the features of final optics in ICF: large amount, large aperture and long distanse, evaluation on health conditions will be conducted before each ICF test. How to acquire promptly the high-definition images of all final optics, how to identifying the defects, locating and measuring the size of defects accurately are viewed as the key technologys in final optics online inspection (FOOI) research.Based on the SG-III facility, the Final optics online inspection technology was studied. The main research contents of the dissertation are stated as follows:The online inspection technology for final optics damage is investigated. An inspection system is designed and the detecting process is presented as well. According to the assignment of final optics in optical module, a zoom telescope optical system with high resolution is designed to meet the testing requirement. Based on the assignment of optical modules in target chamber, the positioning mechanism is designed to adjust the attitude of the imaging module for acquiring images of all the432Final optics.In view of the impact of uneven illumination and “Lighter background” on the accuracy of defects identification, automatic recognition algorithm of defects for FOOI image is explored. With regard to the variation of the defects and its local background, a robust defects detecting method based on Local Area Signal Strength ratio (LASS) and2-D histogram is proposed, theoretically and experimentally. An eliminating method of artificial defects based on AdaBoost algorithm is presented to increase the accuracy of defects identification. In view of the impact of gray saturation, charge overflows and other factors on the precision of defects quantitative identification, a high precision method based on regression is advanced for the quantitative detection of defects. A regression modle is set up via Least Squares Support Vector Machine (LSSVM), and RANSAC is adopted for optimizing the training samples to avoid the effect of outliers.The feasibility and scheme of increasing the detecting resolution based on image mosaic by the proposed online inspection system is studied. In conjunction with the motion parameters of the inspection system, an image registration method based on similarity evaluation of normalized feature is proposed. A global registration method based on4-neighborhood is proposed to reduce the cumulative error caused by serial steps. Poisson image fusion is adopted for eliminating the “stitching line” caused by exposure difference. |
PDF Full Text Request