Micro-optical components defect detection based on light field microscope imaging

Abstract

With the continuous development of cutting-edge manufacturing technology, various micro-nano structures emerge in an endless stream, promoting the development of optical instruments towards the direction of high precision and miniaturization. As an important part of optical instrument, micro-optical components have been widely used in aerospace, medical imaging, industrial inspection and other fields. Their quality directly affects the precision of the whole instrument, so it is of great significance to the quality inspection of micro-optical components. The traditional detection methods are usually manual or mechanical scanning detection based on the optical microscope. Due to the limitation of depth of field of the optical microscope, the detection process requires data acquisition and processing several times, which takes a long time and costs a lot. In recent years, light field imaging technology, as a new method of computational imaging, has been widely used in intelligent terminals, AR/VR, industrial dynamic detection and other fields. Furthermore, light field microscope imaging technology can improve the depth of field of the system and realize multiple imaging with one exposure. In order to overcome the limitations of traditional scanning imaging methods, light field microscope imaging technology is introduced to expand the depth of field range of the system, and one-time detection of micro-optical components can be realized without scanning. But at the same time, light field microscope imaging technology also brings about the problems of low image resolution and low efficiency of digital refocusing calculation, which greatly limits the application of light field microscope imaging in industrial detection. In order to solve the limitation of light field microscope imaging, this paper carried out a detailed study and proposed a complete set of micro-optical components defect detection scheme based on light field microscope imaging and deep learning.