基于云层去除的偏振光导航应用研究  

Research on polarized light navigation application based on cloud removal

作　　者：卢研宏 李沅[1] 屈绍宇 LU Yanhong;LI Yuan;QU Shaoyu(School of Information and Communication Engineering,North University of China,Taiyuan 030051,China)

机构地区：[1]中北大学信息与通信工程学院,山西太原030051

出　　处：《现代电子技术》2024年第21期171-176,共6页Modern Electronics Technique

基　　金：山西省科技成果转化引导专项项目(202204021301044);山西重点研发计划项目(202202010101007)。

摘　　要：传统的导航系统因其自身的技术缺陷而远远不能满足人们日常的导航需求,仿生式偏振光导航应运而生。在传统的图像式偏振光导航系统中,云层遮挡往往会对偏振光信息的采集造成严重影响,从而影响到航向角的准确解算,极大程度限制了偏振光导航的运用范围。为了克服这一困难,文中引入生成对抗神经网络,在生成器与鉴别器之间不断迭代反馈,对采集到的云层图像进行去云层处理,有效地恢复了被云层遮挡的偏振光信息。经过大量实验验证,成功将航向角误差从2.3706°降低至0.6067°,使其导航精度提升了74.41%。实验结果不仅验证了该方法的有效性,也展示了生成对抗网络在图像修复方面的强大之处,对图像式偏振光导航的发展提供了新的思路。The traditional navigation systems are far from being able to meet people's daily navigation needs because of their own technical defects,and bionic polarized light navigation came into being.In the traditional image polarized light navigation systems,cloud occlusion often has a serious impact on the acquisition of polarized light information,thus affecting the accurate calculation of heading angle,which limits the application range of polarized light navigation greatly.In order to overcome this difficulty,a generative adversarial network(GAN)is introduced to remove the cloud from the collected cloud image by iterative feedback between the generator and the discriminator,which can effectively recover the polarized light information blocked by the cloud.After a large number of experiments,the heading angle error is reduced from 2.3706°to 0.6067°successfully,and the navigation accuracy is improved by 74.41%.The experimental result not only verifies the effectiveness of the method,but also shows the power of GAN in image inpainting,so it provides a new idea for image-based polarized light navigation.

关 键 词：仿生式偏振光导航 云层遮挡 生成对抗神经网络 航向角解算 图像修复 导航精度 

分 类 号：TN967.6-34[电子电信—信号与信息处理]