DeepLabv3+与联合损失函数的遥感影像建筑物分割  

Building segmentation in remote sensing image based on DeepLabv3+and joint loss function

作　　者：苏日亚 杨彦明[1] 安全[1] 于建明 SU Ri-ya;YANG Yan-ming;AN Quan;YU Jian-ming(Seismological Bureau of Inner Mongolia Autonomous Region,Hohhot 010010,China)

机构地区：[1]内蒙古自治区地震局,呼和浩特010010

出　　处：《信息技术》2023年第7期38-42,共5页Information Technology

基　　金：中国地震局地震应急青年重点任务(CEA_EDEM-202103)。

摘　　要：快速、自动从遥感影像中提取建筑物可为城市管理、军事侦查、灾后应急评估等提供辅助决策依据。采用基于ResNet50_vd骨干网络的DeepLabv3+深度学习语义分割模型,结合BCE和Lovasz联合损失函数优化算法,实现遥感影像的建筑物提取。在Inria数据集上训练、评估和预测结果显示,采用方法可成功提取遥感影像中的建筑物,准确率最高可达99.02%,mIOU最高可达88.55%。Rapid and automatic extraction of buildings from remote sensing images can provide auxiliary decision-making basis for urban management,military investigation and post disaster emergency assessment.The DeepLabv3+deep learning semantic segmentation model based on ResNet50_vd backbone network is adopted combined with the back propagation optimization model of BCE and Lovasz joint loss function algorithm to realize the building extraction of remote sensing image.The training,evaluation and prediction results on the Inria dataset show that the method can successfully extract buildings from remote sensing images,with the highest accuracy of 99.02%and mIOU of 88.55%.

关 键 词：遥感影像 深度学习 语义分割 建筑物提取 联合损失函数 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]