基于非邻域依赖的实测大气偏振模式重构方法
Reconstruction Method for Measured Atmospheric Polarization Mode Based on Non-Neighborhood Constraint
王文莉 1高欣健 2鲍昌皓 1石磊 1王昕 1高隽1
作者信息
- 1. 合肥工业大学计算机与信息学院,安徽 合肥 230009;合肥工业大学图像信息处理研究室,安徽 合肥 230009
- 2. 合肥工业大学计算机与信息学院,安徽 合肥 230009
- 折叠
摘要
由于获取实测大气偏振数据时往往受地表环境以及天空浮云干扰,大气偏振数据会在不同区域出现大面积失真.针对该问题,提出了一种基于非邻域依赖的实测大气偏振模式重构方法.该方法首先基于同一环境下大气偏振模式分布只与太阳位置有关的特性,通过控制变量法获取实测条件下受干扰大气偏振数据的真实值;其次,基于大气偏振模式在空间上的连续性分布,通过网络的非邻域特征修复模块挖掘大气偏振数据非邻域间的依赖性,以此实现大面积失真下实测大气偏振数据的重构.消融实验的结果表明,当浮云干扰面积大于60%时,在网络的非邻域特征修复模块作用下峰值信噪比和结构性相似指数分别提高了17%和25%.
Abstract
The interference caused by clouds and the surrounding environment leads to significant distortions in the atmospheric polarization data over large areas.We propose an atmospheric polarization mode reconstruction method based on non-neighborhood constraints to address this issue.First,this method leverages the fact that the distribution of atmospheric polarization is solely dependent on the position of the Sun during the same measurement to obtain ground truth data affected by interference through the control variable approach.Second,considering the continuous spatial distribution of atmospheric polarization modes,our approach incorporates a non-neighborhood feature restoration module to uncover dependencies between non-neighborhood regions within the atmospheric polarization data.This capability enables the reconstruction of measured atmospheric polarization data even in case of extensive distortion across large areas.Finally,the results of our ablation experiment show that the indicators peak signal-to-noise ratio and structural similarity are increased by 17%and 25%compared with the non-neighborhood constraint module when the cloud area reaches about 60%.
关键词
大气偏振模式/真实值获取/非邻域特征修复/实测偏振数据重构Key words
atmospheric polarization mode/acquisition of the ground truth/non-neighborhood feature inpainting/reconstruction of measured polarization data引用本文复制引用
基金项目
国家自然科学基金(62171178)
国防科技创新特区项目(11040-41342022001)
出版年
2024
NETL
NSTL
万方数据