计算机技术与发展2026，Vol.36Issue(1)：212-220,211,10.DOI:10.20165/j.cnki.ISSN1673-629X.2025.0309
预训练模型范式迁移视角下的遥感影像技术分析
Analysis of Remote Sensing Image Technology from Perspective of Pre-trained Model Paradigm Shift
摘要
Abstract
Remote sensing image analysis,a pivotal field in geographic information science,has undergone a technological evolution from semi-automated to automated multi-dimensional,high-resolution analytics.Driven by computer vision and deep learning,this field now exhibits diversified methodologies and increasingly domain-specialized approaches.We systematically analyze 58 scholarly publications,focusing on three key research directions:scene classification,image retrieval,and image segmentation.It comparatively examines core datasets,methodological distinctions,and performance evaluation systems across dimensions such as feature extraction and semantic analysis.Key findings reveal two critical research bottlenecks:inefficient multi-modal data fusion and limited task integration.To address these challenges,we propose developing open-source sharing platforms,optimizing large model-driven feature extraction accuracy,and strengthening multi-task collaborative frameworks.Not only does this research fill the gap in meta-analytical studies of remote sensing imagery within the GIS domain,but it also provides theoretical foundations and optimization pathways for future technological iterations.关键词
遥感影像分析/场景分类/特征提取/语义分析/图像检索/目标分割/荟萃分析Key words
remote sensing image analysis/scene classification/feature extraction/semantic analysis/image retrieval/object segmentation/meta-analysis分类
信息技术与安全科学引用本文复制引用
李冰,侯锐,杨晓艳,石涛,孟令通,赵琛浩..预训练模型范式迁移视角下的遥感影像技术分析[J].计算机技术与发展,2026,36(1):212-220,211,10.基金项目
中国科学院战略性先导科技专项(XDA15040200) （XDA15040200）
