面向多源数据的多区域尺度协同高分遥感图像语义分割  

Multi Region-scale Collaborative Semantic Segmentation of High-resolution Remote Sensing Images for Multi-source Data

作　　者：林易丰 陈光剑 陈浩[1,2] 翁谦[1,2,3] 林嘉雯[1,2,3] LIN Yifeng;CHEN Guangjian;CHEN Hao;WENG Qian;LIN Jiawen(College of Computer and Data Science,Fuzhou University,Fuzhou 350116,China;Fujian Key Laboratory of Network Computing and Intelligent Information Processing,Fuzhou University,Fuzhou 350116,China;Key Lab of Spatial Data Mining and Information Sharing,Ministry of Education,Fuzhou University,Fuzhou 350116,China;Fujian Geologic Surveying and Mapping Institute,Fuzhou 350116,China)

机构地区：[1]福州大学计算机与大数据学院,福州350116 [2]福州大学福建省网络计算与智能信息处理重点实验室,福州350116 [3]福州大学空间数据挖掘与信息共享教育部重点实验室,福州350116 [4]福建省地质测绘院,福州350116

出　　处：《小型微型计算机系统》2025年第1期158-166,共9页Journal of Chinese Computer Systems

基　　金：福建省自然科学基金项目(2023J01432,2020J01230079)资助;福厦泉国家自主创新示范区协同创新平台项目(2022FX5)资助;国家自然科学基金项目(41801324)资助。

摘　　要：在高分遥感图像语义分割中,为解决如何有效融合光谱信息与高程信息以分割相似光谱的不同地物的问题和通过捕获长距离依赖信息来提升局部地物识别精度,本文提出一种面向多源数据的多区域尺度协同语义分割方法.该方法包括:一种不等长的多分支语义分割网络,以有效提取多源特征,充分利用多源数据之间的互补信息;一个轻量级的协同注意力特征融合模块,用于在特征融合阶段有效地融合多分支特征;一种多区域尺度协同的数据增强方法,引导网络捕获长距离依赖信息.在ISPRS提供的公开数据集Vaihingen和Potsdam上的实验结果表明,与同类型主流方法对比,本文提出的方法具有更优的分割性能,且得到的地物细节信息更加完整,参数量也更小.In the semantic segmentation of high-resolution remote sensing images,in order to solve the problem of how to effectively fuse spectral and elevation information to segment different features with similar spectra and to improve the local feature recognition accuracy by capturing long-range dependent information,this paper proposes a multi-region scale collaborative semantic segmentation method for multi-source data.The method includes:an unequal-length multi-branch semantic segmentation network to efficiently extract multi-source features and fully utilize the complementary information between multi-source data;a lightweight collaborative attentional feature fusion module for efficiently fusing multi-branch features in the feature fusion stage;and a multi-region scale collaborative data enhancement method that guides the network to capture long-range dependent information.Experimental results on the publicly available datasets Vaihingen and Potsdam provided by ISPRS show that the method proposed in this paper has a better segmentation performance compared with mainstream methods of the same type,and yields more complete information on feature details and a smaller number of parameters.

关 键 词：语义分割 高分遥感图像 数字表面模型 多源数据融合 协同注意力 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]