基于渐进式生成对抗网络的舰船红外图像仿真

红外图像仿真在红外导引头设计、仿真训练中起到十分关键的作用。针对如何生成高分辨率、视觉特征可控的红外图像,提出了一种基于渐进式生成对抗网络的红外图像仿真方法。本文利用舰船模型的红外图像数据集训练了图像合成网络,输入随机特征向量,输出高分辨率的红外仿真图像;设计了图像编码网络,实现红外图像到特征向量的转换;利用Logistic回归方法,在特征向量域找到了控制红外图像角度特征的方向向量,并据此生成了不同角度的舰船模型仿真图像;最后通过均值哈希算法和平均结构相似性算法来定量评价仿真图像和真实图像的差异,实验结果表明仿真的红外图像和真实图像的相似度很高,可以为真实舰船的可控化红外图像仿真提供参考。

Infrared image simulation for warship based on progressive generative adversarial network

Infrared image simulation plays an important role in the design and training of infrared seeker.Aiming at how to simulate infrared images with high resolution and controllable visual features,a method of infrared image simulation based on progressive generative adversarial network is proposed.In this paper,the infrared images dataset of warship model is used to train the image synthesis network,only random feature vector is necessary for the generation of high resolution infrared simulated image.An image encoder network is designed to convert infrared images to feature vectors.By the logistic regression method,the direction vector which controls the angle feature of infrared images is found in the feature vector domain,and simulated images of warship model from different angles are generated accordingly.AHash algorithm and MSSIM algorithm are used to quantitatively evaluate the difference between simulated and real images.The experimental results show that the similarity between the simulated infrared images and the real ones is very high,and it can be used as a reference for the controllable simulation of real warship infrared images.