机器学习模型在地质灾害遥感调查数据分析中的应用现状

Current application of machine learning models in the analysis of remote sensing survey data for geological hazards

为研究机器学习模型在地质灾害遥感调查中的应用现状,基于中国知网(CNKI)数据库,采用文献计量法进行可视化分析,从发文量、研究热点、研究机构等多视角详述机器学习模型、地质灾害遥感调查技术的研究进展。利用VOSviewer软件分析机器学习模型与地质灾害遥感调查技术高频关键词及其关联度,并通过分类统计定量化分析得出研究热点、关联性和发展趋势。结果表明:中国地质灾害遥感调查技术正由“图谱测量”向“图谱与几何测量”逐步转变,新一代机器学习算法伴随着无人机遥感技术的进步,已成为本领域最热门的研究方向,推动着地质灾害体自动识别和智能提取技术发展;未来的地质灾害遥感调查技术必然是围绕“空−天−地”协同应用与应急监测的综合技术体系。研究认为,针对不同遥感影像数据的特点,综合研究不同机器学习模型在各种遥感解译工作场景中的应用是未来的主要发展趋势。

To investigate the current landscape of the application of machine learning in remote sensing surveys of geological disasters and to support the development of intelligent remote sensing survey technologies for geological disasters,a bibliometric analysis of machine learning and geological disaster remote sensing survey technology was conducted using the China National Knowledge Infrastructure(CNKI)database.Visual analysis was performed from multiple perspectives,including the number of publications,research hotspots,and research institutions,to describe the research progress of machine learning and geological disaster remote sensing survey technology.VOSviewer software was utilized to scrutinize the high-frequency keywords and their associations between machine learning and geological disaster remote sensing survey technology.The results showed that remote sensing survey technology for geological disasters in China is gradually shifting from traditional“topographic measurement”towards more holistic“topographic and geometric measuremen”approaches.With the advancement of unmanned aerial vehicle remote sensing technology,the new generation of intelligent learning algorithms have emerged as the predominant research direction,fostering the growth of automated geological disaster recognition and intelligent extraction techniques.Nevertheless,the future of remote sensing survey technology for geological disasters is poised to evolve into a comprehensive technical system that emphasizes the synergistic“air-space-ground”application and emergency monitoring.Considering the diverse characteristics of remote sensing image data,the primary developmental trajectory will involve an extensive exploration of various machine learning algorithms across different remote sensing interpretation scenarios.