从图像到图像的场景重建技术研究

Research on digital camouflage scheme based on pix2pix

针对夜间伪装过程中目标与背景颜色难以融合的问题,开展了基于pix2pix网络进行场景重建技术的研究,利用该网络强大的学习能力,学习不同温度下特定的颜色分布和背景纹理特征并预测颜色信息,实现红外图像与彩色图像之间的转换,在pix2pix框架下生成逼真的实景图案。无论在数学上还是实验上,相较现有的夜间场景重建技术,生成的场景图案与真实背景相似度明显提高,解决了夜间能见度不足情况下迷彩伪装的设计难题。

Aiming at the problem that it is difficult to fuse the target and the background color in the process of night camouflage,the scene reconstruction technology based on the pix2pix network is proposed in this paper.The strong image-to-image learning ability of this network is used to learn the specific color distribution and background texture features in a certain area under different temperatures,and predict the color information.The conversion between infrared heat map and color image is realized,and a realistic real scene pattern is generated under the framework of pix2pix.In both mathematical and experimental terms,compared with the existing reconstruction techniques of night scenes,the fusion degree of scene patterns generated in this paper and the real background is significantly improved,which solves the design problem of camouflage camouflage under the condition of insufficient visibility at night.