基于图像显著性的空中光学伪装效果评估方法  

Evaluation Method for Aerial Optical Camouflage Effect Based on Image Saliency

作　　者：张祥 沙建军 郭铭洋 黄倩 温远征 Zhang Xiang;Sha Jianjun;Guo Mingyang;Huang Qian;Wen Yuanzheng(College of Intelligent Systems Science and Engineering,Harbin Engineering University,Harbin 150001,China;Qingdao Innovation and Development Base,Harbin Engineering University,Qingdao 266000,China;College of Information and Communication Engineering,Harbin Engineering University,Harbin 150001,China)

机构地区：[1]哈尔滨工程大学智能科学与工程学院,哈尔滨150001 [2]哈尔滨工程大学青岛创新发展基地,青岛266000 [3]哈尔滨工程大学信息与通信工程学院,哈尔滨150001

出　　处：《兵工自动化》2024年第9期7-12,共6页Ordnance Industry Automation

摘　　要：针对传统伪装效果评估在大场景空中图像不适用问题,提出一种基于图像显著性的空中光学伪装效果评估方法。采用频率调谐(frequency-tuned,FT)模型生成全局显著图和显著区域图,使用残余谱(spectrum residual,SR)模型生成局部显著图;根据显著图计算全局显著均值、显著区域占比和局部显著均值3个指标;通过熵权法综合3个指标得到综合显著性评估系数来量化评估,在林地和雪地背景下获取伪装目标空中光学图像进行实验验证。结果表明:该方法能较好地体现不同伪装效果的差异,优于目标背景相似度加权模型,且可根据显著图优化伪装方案。In response to the problem that traditional camouflage effect evaluation is not applicable to large-scale aerial images,a method for evaluating aerial optical camouflage effect based on image saliency is proposed.Using frequency tuned(FT)models to generate global saliency maps and saliency region maps,and using residual spectrum(SR)models to generate local saliency maps;Calculate three indicators:global significance mean,proportion of significant regions,and local significance mean based on saliency maps;By using the entropy weight method to synthesize three indicators,a comprehensive significance evaluation coefficient is obtained to quantify the evaluation.Optical images of disguised targets in the air are obtained under forest and snow backgrounds for experimental verification.The results show that this method can effectively reflect the differences in camouflage effects,outperform the target background similarity weighted model,and can optimize camouflage schemes based on saliency maps.

关 键 词：图像显著性 熵权法 伪装评估 空中光学图像 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]