检索规则说明:AND代表“并且”;OR代表“或者”;NOT代表“不包含”;(注意必须大写,运算符两边需空一格)
检 索 范 例 :范例一: (K=图书馆学 OR K=情报学) AND A=范并思 范例二:J=计算机应用与软件 AND (U=C++ OR U=Basic) NOT M=Visual
作 者:刘彦德 王竞雪[1,2] LIU Yande;WANG Jingxue(School of Geomatics,Liaoning Technical University,Fuxin,Liaoning 123000,China;Faculty of Geosciences and Environmental Engineering,Southwest Jiaotong University,Chengdu 611756,China)
机构地区:[1]辽宁工程技术大学测绘与地理科学学院,辽宁阜新123000 [2]西南交通大学地球科学与环境工程学院,成都611756
出 处:《遥感信息》2023年第2期96-103,共8页Remote Sensing Information
基 金:国家自然科学基金项目(41871379);辽宁省兴辽英才计划项目(XLYC2007026);辽宁省应用基础研究计划项目(2022JH2/101300273)。
摘 要:针对传统的水体指数法和监督分类法在对卫星影像进行水体提取时经常出现错分的问题,提出一种改进Swin Transformer网络的资源三号影像水体提取方法,并对其特征金字塔的构建进行优化。其中,结合PANet算法原理快速融合特征金字塔顶层信息和底层信息,通过融合不同尺度的信息提高对水体特征提取的效果;利用Swin Transformer网络包含的滑动窗口分割和循环位移操作来减少训练时的计算复杂度。通过混淆矩阵计算得到该方法对城区、郊区和山区水体提取结果的总体精度、Kappa系数、F1-score指数,分别在90.14%~96.52%之间、0.80~0.88之间、0.81~0.87之间,证明该方法对资源三号卫星影像进行水体提取的有效性与可靠性。Aiming at the problem that the traditional water body index method and supervised classification method often causes misclassification when extracting water bodies from satellite images,this paper proposes an improved Swin Transformer network for the water body extraction method of ZY-3 image.The building of its feature pyramid is optimized in the method.Combined with the principle of the PANet algorithm,the top-level and bottom-level information of the feature pyramid are quickly fused,and the effect of water feature extraction is improved by fusing information of different scales.The computational complexity is reduced by using the sliding window segmentation and cyclic displacement operations included in the Swin Transformer network during training.In the experiment,the overall accuracy,Kappa coefficient,and F1-score index for urban,suburban and mountainous water extraction results are obtained by calculating the confusion matrix.The corresponding figure ranges from 90.14%to 96.52%,from 0.80 to 0.88,and from 0.81 to 0.87,respectively.The results of the experiment show that the method for water extraction from ZY-3 satellite images is effective and reliable.
关 键 词:Swin Transformer PANet 资源三号卫星 水体提取 Kappa系数
分 类 号:TP751[自动化与计算机技术—检测技术与自动化装置]
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在链接到云南高校图书馆文献保障联盟下载...
云南高校图书馆联盟文献共享服务平台 版权所有©
您的IP:216.73.216.49