可抵抗团雾的无人机航拍图像去雾算法  

Defogging Algorithm for Aerial Image of UAV Can Resist Fog

作　　者：韩丽 黄冠 王沛祺 刘岩 HAN li;HUANG Guan;WANG Pei-qi;LIU Yan(College of Computer and Communication Engineering,Zhengzhou University of Light Industry,Zhengzhou Henan 450001,China;Cyberspace Security College,Zhengzhou University,Zhengzhou Henan 450001,China)

机构地区：[1]郑州轻工业大学计算机与通信工程学院,河南郑州450001 [2]郑州大学网络空间安全学院,河南郑州450001

出　　处：《计算机仿真》2024年第11期205-209,219,共6页Computer Simulation

摘　　要：无人机航拍技术的进步推进无人机航拍高度和视场不断增加,视场和拍摄高度的提高导致雾天环境下更易捕获团雾图像。现有去雾算法大多针对图像传感器低视角下所捕获的均匀雾图进行处理,而对无人机在高度视角下所捕获的团雾图像去雾效果较差。提出一种可抵抗团雾影响的无人机航拍图像去雾算法(MLD-Net)。通过引入匀光模块对图像团雾区域特征进行均匀性调整,增强图像因团雾遮挡导致的局部过暗所带来的细节损失,提升图像质量。提出的MLD-Net网络输出图像在信息熵、平均梯度等评价指标表现优于现有研究成果,输出图像细节特征丰富,更易于后期高级算法处理。所提出算法不仅适用于非均匀团雾情况下,同样适用于薄雾和浓雾情况下,航拍图像去雾效果显著。The progress of UAV aerial photography technology promotes the increase of UAV aerial photography height and field of view.The improvement of the field of view and shooting height makes it easier to capture the fog image in a foggy environment.Most of the existing dehazing algorithms deal with the uniform fog image captured by the image sensor at a low angle of view,while the fog removal effect of the fog image captured by UAV at a high angle of view is poor.This paper proposes an unmanned aerial vehicle aerial image dehazing algorithm(MLD-Net)that can resist the influence of cloud fog.The uniform light module is introduced to adjust the uniformity of the features of the fog area of the image,enhance the detail loss caused by the local over-dark caused by the fog occlusion,and improve the image quality.The output image of the MLD-Net network presented in this paper outperforms the existing research results in terms of information entropy,average gradient and other evaluation indicators.The output image has rich detail features and is easier to be processed by advanced algorithms in the later stage.The proposed algorithm is applicable not only to the case of non-uniform fog,but also to the case of thin fog and thick fog.

关 键 词：无人机航拍 团雾图像 去雾算法 信息熵 平均梯度 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]