Abstract
The main influencing factor of imaging under high speed conditions is the aero-optical effect. Aero-optical effect is a kind of noise, which can be regarded as a superposition of three kinds of noises, system noise ,aero transmission effect and aero heat radiation effect. In this paper, we only consider aero transmission effect and aero heat radiation effect. The simulation and correction of aero-optical effects are important for terminal guidance. The aero-optical effect simulation method described in this paper uses a deep neural network (Conditional Generative Adversarial Networks). The use of big data allows the conditional generative adversarial networks to learn the mapping between clear pictures and pictures with aero-optical effects in training. Aero-optical effect correction is usually divided into two parts, the correction of the aero transmission effect and the correction of the aero heat radiation effect. In this paper, we use the conditional generative adversarial networks to train a large amount of data and learn the mapping relationship between the pictures with the aero-optical effects and the clear pictures in the training. And this mapping relationship is preserved in the form of bias and weights. It is not necessary to consider the aero-optical effect and the aero heat radiation effect separately. In this paper, the structural similarity (SSIM) between the real image and the simulated image generated by the conditional generation adversarial the network is 97.63%. The structural similarity (SSIM) between the clear image and the aero-optical effect image corrected with the conditional generation adversarial the network is 76.59%. The structural similarity (SSIM) between the original aero-optical effect image and the clear image is 55.73%.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have