一种融合场景解译与面向对象分割精准识别城市边缘区域闲置地的方法  

Method combining scene interpretation and object-oriented segmentation to accurately identify idle land in urban fringe areas

作　　者：谢强 XIE Qiang(Jiangsu Geologic Surveying and Institute,Nanjing 211102,China)

机构地区：[1]江苏省地质测绘院,江苏南京211102

出　　处：《江苏科技信息》2024年第14期127-130,136,共5页Jiangsu Science and Technology Information

摘　　要：为了快速准确识别城市边缘地区闲置土地,文章基于高分2号遥感影像提出了一种融合场景解译与面向对象分割的方法,用于识别城市边缘区域的闲置土地。文章引入四叉树分割方法制作了多尺度场景解译标签;选取性能优越的Vision Transformer模型作为提取工具,进行训练以得到解译模型;将该解译模型应用于经过面向对象分割处理后的影像对象进行场景分类。实验结果表明,相比传统方法,该方法在闲置土地识别方面取得了显著提升。The present study proposes a method based on Gaofen-2 remote sensing image data to efficiently and accurately identify idle land in the urban fringe by integrating scene interpretation and object-oriented segmentation.Initially,the quadtree segmentation technique is employed to generate multi-scale scene interpretation labels.Subsequently,the superior Vision Transformer model is selected as the extraction tool for training purposes to acquire an interpretation model.Finally,this interpretation model is applied to object-oriented image objects after undergoing object segmentation for scene classification.Experimental results demonstrate that compared with conventional approaches,this method exhibits significant advancements in identifying idle land.

关 键 词：城市边缘区域闲置地 场景解译 面向对象分割 遥感影像 深度学习 

分 类 号：P258[天文地球—测绘科学与技术]