The high-resolution detecting system based on machine vision for defects on large aperture and super-smooth surface uses a novel ring telecentric lighting optical system detecting the defects on the sample all round and without blind spots. The scattering light induced by surface defects enters the adaptive and highly zoom microscopic scattering dark-field imaging system for defect detecting and then forms digital images. Sub-aperture microscopic scanning sampling and fast stitching on the surface is realized by using precise multi-axis shifting guided scanning system and a standard comparison board based upon binary optics is used to implement fast calibration of micron-dimension defects detected actually. The pattern recognition technology of digital image processing which can automatically output digitalized surface defects statements after scaling is established to comprehensively evaluate defects. This system which can reach micron-dimension defect resolution can achieve detections of large aperture components of 850 mm × 500 mm, solve the durable problem of subjective uncertainty brought in by human visual detection of defects and achieve quantitative detection of defects with machine vision.
In the inertial confinement fusion system (ICF), surface scratches of the large diameter optical surface appear as dot lines (punctate scratches). This kind of scratches is only detected under a high microscope magnification system. This can be caused by the blemishes on the optical processing technology and shallow scratches (< 25nm ). As a result, it can have an impact on the relevant calculation of the width and length of the scratches. Besides, this kind of scratches has a serious impact on the ICF, such as system damage. To solve this problem, this paper proposes the image pattern charter of punctate scratches based on the existing surface defects detection system (SDES). Finally, it proposes an algorithm of scratches based on the linearity differential detection and connectivity. That is, using coordinate transformation and direction differential-threshold discrimination, the scratches can be connected effectively and calculated exactly. Experimental results show that punctate scratches parts can be connected correctly, and the accuracy of the calculated length reaches 95%. Also, the improved algorithm applies to the arc-shaped scratches, which is based the block image processing. Currently, this algorithm can be applied to connect and calculate the shallow scratches accurately and precisely on large fine optics in the ICF system. Thus it can also decrease the misdetection rate of nonconforming super-smooth optics in the ICF system.