Tolerance desensitization of long-wave infrared systems based on deep learning

Manufacturing and assembly tolerances degrade the imaging quality of optical systems, and image processing techniques can alleviate this effect, thereby reducing manufacturing and assembly requirements and costs. This paper presents an end-to-end optical system image quality enhancement methodology designed to reduce the impact of tolerances on imaging quality. The methodology is developed by constructing the Point Spread Functions (PSFs) grid to generate the simulated image dataset and trained using Generative Adversarial Network to improve imaging quality and reduce the system’s manufacturing and assembly requirements. Both simulation and experimental results validate the effectiveness of the method.