测绘学报 ›› 2017, Vol. 46 ›› Issue (7): 910-917.doi: 10.11947/j.AGCS.2017.20170035

• 摄影测量学与遥感 • 上一篇    下一篇

图割能量驱动的高分辨率遥感影像震害损毁建筑物检测

刘莹, 陶超, 闫培, 邹峥嵘   

  1. 中南大学有色金属成矿预测与地质环境监测教育部重点实验室, 地球科学与信息物理学院, 湖南 长沙 410083
  • 收稿日期:2017-01-18 修回日期:2017-06-20 出版日期:2017-07-20 发布日期:2017-07-25
  • 通讯作者: 陶超 E-mail:kingtaochao@126.com
  • 作者简介:刘莹(1992—),女,硕士,研究方向为高分遥感影像变化检测。
  • 基金资助:
    国家自然科学基金(41301453);湖南省自然科学基金(2017JJ3378);国家973计划(2012CB719903);教育部博士点基金(20130162120027)

Graph Cut Energy Driven Earthquake-damaged Building Detection from High-resolution Remote Sensing Images

LIU Ying, TAO Chao, YAN Pei, ZOU Zhengrong   

  1. Key Laboratory of Metallogenic Prediction of Nonferrous Metals and Geological Environment Monitoring(Central South University), Ministry of Education, School of Geosciences and Info-Physics, Changsha 410083, China
  • Received:2017-01-18 Revised:2017-06-20 Online:2017-07-20 Published:2017-07-25
  • Supported by:
    The National Natural Science Foundation of China (No.41301453);The National Natural Science Foundation of Hunan province, China (No. 2017JJ3378);The National Basic Research Program of China (973 Program) (No.2012CB719903);Research Fund for the Doctoral Program of Higher Education (No.20130162120027)

摘要: 为充分利用高分辨率遥感影像提供的细节信息,提高震害损毁建筑物检测精度,提出了一种图割框架下融合形状、边缘、角点等多种类型特征的损毁建筑物检测方法。该方法首先利用震前建筑物线划图提取包含单个建筑物的局部影像,用于图割能量函数建模,并分别根据建筑物的位置、形状、边缘以及角点构造能量函数的各约束项。在此基础上,通过最大流/最小割算法求解能量函数最小值,依据最小割能量进行地震前后局部影像中建筑物的相似性度量。最后利用最大期望算法(expectation maximization,EM)求得最小割能量的分类阈值,并根据错分率估值执行后处理以获得最终变化检测结果。采用2011年3月11日东日本大地震前后石卷港的高分辨率遥感影像进行试验,结果表明本文提出的方法能有效检测出损毁建筑物。

关键词: 损毁检测, 建筑物, 图割, 高分辨率遥感影像, 面向对象

Abstract: In order to make full use of the detail information provided by high-resolution remote sensing images to improve the detection accuracy of damaged buildings during earthquake,combining the shape, edge and corner characteristics of buildings, a novel method based on graph cut frame for damaged building detection is proposed in this paper. Firstly, the local image containing single building used to model the energy function is constructed by digital line graphic data. And the bound terms of the energy function are defined by the location, shape, edge and corner of buildings, respectively. Then, the energy function is minimized through max-flow/min-cut method, and the similarity of the buildings in the pre-and post-event images is measured by the minimum cut energy. Finally, the EM algorithm is exploited to select the classification threshold value of the minimum cut energy, and post-processing is performed according to the misclassification rate estimation to obtain the final detection result. Images taken in Ishinomaki before and after the 2011 off the Pacific coast of Tohoku Earthquake are used in this paper. The experimental results show that the proposed method can effectively detect the damaged buildings.

Key words: damage detection, building, graph cut, high-resolution remote sensing image, object-oriented

中图分类号: