基于改进金字塔结构的遥感影像建筑物变化检测  

Research on Building Change Detection in Remote Sensing Image Based on improved pyramid structure

作　　者：尉樱樊 刘超[1] 刘春阳[1] WEI Yingfan;LIU Chao;LIU Chunyang(School of Spatial Information and Geomatics Engineering,Anhui University of Science and Technology,Huainan,Anhui 232001,China)

机构地区：[1]安徽理工大学空间信息与测绘工程学院,安徽淮南232001

出　　处：《九江学院学报(自然科学版)》2024年第4期60-64,69,共6页Journal of Jiujiang University:Natural Science Edition

基　　金：国家自然科学基金面上项目(编号42071384);安徽省教育厅科学研究重点项目(编号2022AH050849)的研究成果之一。

摘　　要：针对遥感影像建筑物变化检测任务中存在的漏检、错检和建筑物边缘不完整的问题,文章提出一种基于金字塔结构改进的IFPNet网络,在FPN网络模型的基础上引入scSE注意力模块,同时将编码器替换为Inception V4对小目标进行精细化的提取。在实验条件相同的条件下选择UNet、UNet++、FPN、DeepLab V3+与文章的网络模型进行对比,实验结果表明改进的方法在召回率、IoU和F1三个指标上都取得了最优结果,与FPN相比召回率提升了8.61%,IoU提升了6.98%,F1值提升了4.45%,最终结果证明文章网络模型在进行建筑物变化检测任务时的优越性。Aiming at the problems of leakagewrong detection and incomplete building edges in the task of building change detection in remote sensing images this paper proposed an improved IFPNet network based on pyramid structurewhich introduces the SCSE attention module on the basis of the FPN network model and at the same time replaces the encoder with Inception V4 to refine the extraction of small targets.UNetUNet++FPNDeepLab V3+were selected to compare with the network model of this paper under the same experimental conditionsand the experimental results showed that the improved method achieved the optimal results in the three indexes of recall IoU and F1 and the recall is improved by 8.61%the IOU was improved by 6.98%and the value of F1 was improved by 4.45%compared with that of FPN.The final result proved the superiority of the network model in this paper when performing the building change detection task.

关 键 词：建筑物变化检测 IFPNet网络 小建筑物 

分 类 号：P237[天文地球—摄影测量与遥感]