基于深度学习的海上船舶遥感识别方法对比分析  

Comparative Analysis of Marine Ships Remote Sensing Identification Methods Based on Deep Learning

作　　者：陈秋 邵长高[1,2] 吕建军 CHEN Qiu;SHAO Changgao;LYU Jianjun(Sanya Institute of South China Sea Geology,Guangzhou Marine Geological Survey,Sanya 572025,China;Academy of South China Sea Geological Science,China Geological Survey,Sanya 572025,China;School of Geography and Information Engineering,China University of Geosciences,Wuhan 430078,China)

机构地区：[1]广州海洋地质调查局三亚南海地质研究所,海南三亚572025 [2]中国地质调查局南海地质科学院,海南三亚572025 [3]中国地质大学地理与信息工程学院,湖北武汉430078

出　　处：《地理空间信息》2024年第12期74-78,共5页Geospatial Information

基　　金：三亚崖州湾科技城管理局2022年度科技计划资助项目(SKJC-2022-01-001);海域监测应用资助项目(2022R-SYS25-03)。

摘　　要：在计算机视觉领域,随着深度学习算法不断发展,目标识别模型的性能得以显著提高,但是针对高分辨率遥感影像中的小目标识别仍具有挑战性,特别是海上小型作业船舶的识别难度较高,而船舶识别在安全、交通、军事等领域重要性日渐突出。基于此,对用于目标识别的经典数据集以及用于船舶识别的数据集进行了介绍,详细分析了目标识别的模型的特征及优缺点,并进行了对比分析,展望了深度学习在船舶遥感识别的未来研究方法,提出多模态融合技术的应用。The continuous development of deep learning algorithms let the ability and efficiency of target identification models improved significantly in the study field of computer vision.However,the identification of small targets in high-resolution remote sensing images remains challenging,especially for small operating ships.With the development of marine economic,the identification of ships becoming more and more important in different areas,such as maritime safety,transportation and military.We discussed the classic data sets used for target identification and ship identification in the present study,studied and contrastive analyzed the characteristics,advantages and disadvantages of target identification model in detail,prospected the future research methods of deep learning in ship remote sensing identification,and put forward the application of multi-modal fusion technology.

关 键 词：深度学习 遥感影像 船舶 船舶识别 

分 类 号：P237[天文地球—摄影测量与遥感]