Use Deep Residual GAN to Inference 3D Space Thin Film Laser Damage

Laser bombardment damage to thin films is one of the more important topics in optical instrumentation. The distribution and morphology of the damaged spot can be studied to measure the directionality and energy stability of the laser. It can also be used to examine the uniformity and purity of the coating, which is a mutually verifiable experiment. After the laser has finished bombarding, the traditional means is to use interference microscopy to obtain the depth matrix of the damaged area, and many scholars and research institutions have used this method. However, compared to electron microscopy, high magnification optical cameras have a more significant advantage in portability and cost. If a way can obtain the spatial depth matrix of the damaged spot using only an optical camera, it will significantly reduce the cost of research in the discipline. This paper proposes a deep residual generative adversarial neural network to infer damage depth matrix using CCD images. It is found that although our model still has some room for improvement, this may be a bold attempt.