用于夜间撒布地雷探测的可见光偏振图像融合  

Visible-polarized image fusion for nighttime dispersal of mines

作　　者：沈康宇 崔博伦 吕其峰 王驰 SHEN Kangyu;CUI Bolun;LYU Qifeng;WANG Chi(School of Mechanical Engineering and Automation,Shanghai University,Shanghai 200444,China;Beijing Institute of Space Mechanics&Electricity,Beijing 100094,China)

机构地区：[1]上海大学机电工程与自动化学院,上海200444 [2]北京空间机电研究所,北京100094

出　　处：《光学精密工程》2024年第15期2439-2453,共15页Optics and Precision Engineering

基　　金：北京市航空智能遥感装备工程技术研究中心开放基金课题(No.AIRSE20233);国家重点研发计划资助项目(No.2023YFF0722902);国家自然科学基金资助项目(No.62175144)。

摘　　要：为了克服在夜间光线微弱情况下,撒布地雷目标与周边地面背景的光谱强度差异较弱的难题,研究一种端到端的无监督可见光偏振图像融合增强算法。融合图像利用撒布地雷的偏振特性在夜间增强地雷目标的同时,尽可能地保留场景的细节纹理信息。融合算法网络由特征提取模块、特征融合模块和图像重构模块构成。在特征融合方面,引入混合注意力机制以加强网络对特征张量中显著信息的提取能力。并通过设计基于像素内容分布的损失函数,引导融合图像保留更多源图像中显著像素区域特征,实现网络的端到端输出。针对夜间撒布地雷数据集,同7种主流图像融合方法进行主客观评价,并在SSIM,VIF等8项评价指标中表现为同类最优,YOLOv5模型中经融合增强的图像在地雷检测任务中表现要优于强度图像(mAP@0.5领先8%,mAP@0.5:0.95领先11%)。本模型具备先进性,并对后续地雷目标检测任务有积极影响。To address the challenge of weak spectral intensity differences between dispersed mine targets and the surrounding ground in low light conditions at night,an end-to-end unsupervised visible-polarized image fusion enhancement algorithm is explored.This algorithm uses the polarization characteristics of scattered mines to enhance nighttime mine targets while preserving scene texture details.The fusion algorithm network consists of a feature extraction module,a feature fusion module,and an image reconstruction module.A hybrid attention mechanism is incorporated to improve the network's ability to extract significant information from the feature tensor.Additionally,a loss function based on pixel content distribution is designed to ensure the fused image retains prominent pixel features from the source image,enabling end-to-end network output.For the nighttime landmine scattering dataset,evaluations using seven mainstream image fusion methods showed superior performance across eight metrics,including SSIM and VIF.The fusion-enhanced image in the YOLOv5 model surpassed the intensity image in landmine detection tasks.This model is state-of-the-art and positively impacts subsequent mine detection missions.

关 键 词：图像融合 撒布地雷 偏振特性 可见光 低照度 

分 类 号：TN209[电子电信—物理电子学] O436.3[机械工程—光学工程]