Research On Automatic Detection And Repair System Of Surface Micro Defects On Aspheric Optics

Aspheric optics are the key components of the terminal optical components of high-power solid-state laser devices.Under the strong laser irradiation,the surface of the component is prone to micro-defects.Once the defects are generated,the damage process will be exacerbated,and at the same time,the downstream components will be damaged.Due to the large number and the small size of defects,manual operation to detect and repair is inefficient and error-prone.Therefore,it is of great significance to automate the entire process of detecting and repairing micro-defects on the component surface.To achieve this goal,based on the CO₂ single point laser repair prototype,the design of automatic detection and repair system of surface micro-defects,image processing algorithm of automatic detection and automatic repair scheme of micro-defects on the surface are studied in this thesis.The original prototype station is transformed according to the automatic function requirements of defect detection and repair.Corresponding automated process flow is designed for the modified work station layout,including:automatic determination of component surface position,automated dark field inspection,automated bright field microscopic inspection,and automated repair and repair effect records.In addition,this thesis also developed corresponding automation control software for the detection and repair system.The realization of the automated detection process depends on the extraction of image information.This thesis has studied the image processing algorithms involved in this process.Aiming at the edge detection of components,an object distance focusing method based on the variation curve of gray scale is proposed to obtain a clear image of the edge,and the accurate position of the edge is determined by the method of gradient operator combined with binarization.Aiming at the automated dark field detection process,a defect extraction method of top-hat transformation combined with adaptive threshold segmentation is proposed to achieve the acquisition of the coarse positioning coordinates of the defect bright field.An automatic focusing algorithm for defect brightfield microscopic images and a minimum circumscribed circle extraction algorithm for defect targets are designed to realize the precise positioning process of defects.In addition,this subject will also apply convolutional neural networks to defect detection and recognition.This thesis has designed an image segmentation model based on encoder-decoder structure to achieve defect target extraction,and used Mobile Net-V2 neural network structure to remove pseudo-defects.In order to realize the automatic repair of micro-defects on the surface of the component,this thesis has studied the laser repair strategy,the automatic control of the laser repair system and the record of the laser repair effect.According to the size information of the defects and the relative positional relationship,three types of defect repair strategies,including single defect point repair,multi-defect points repair and manual decision-making,are set up,and a repair plan file is generated.The hardware composition of the laser repair system is analyzed,and the corresponding software control system is developed to realize automatic control of the laser parameters.For the repair pits of different sizes,the bright field micro camera single-frame photo and 3×3scanning photo programs are designed to achieve automatic recording of the repair effect.