地球信息科学学报 ›› 2017, Vol. 19 ›› Issue (5): 692-701.doi: 10.3724/SP.J.1047.2017.00692

• 遥感科学与应用技术 • 上一篇    下一篇

无人机遥感数据处理与滑坡信息提取

陈天博(), 胡卓玮*(), 魏铼, 胡顺强   

  1. 1. 首都师范大学资源环境与旅游学院,北京 100048;
    2. 首都师范大学 资源环境与地理信息系统北京市重点实验室,北京 100048;
    3. 首都师范大学 三维信息获取与应用教育部重点实验室,北京 100048
  • 收稿日期:2016-01-21 修回日期:2016-06-23 出版日期:2017-05-20 发布日期:2017-05-27
  • 通讯作者: 胡卓玮 E-mail:chen.tianbo@163.com;huzhuowei@gmail.com
  • 作者简介:

    作者简介:陈天博(1990-),男,硕士生,主要从事无人机遥感影像处理与信息提取研究。E-mail:chen.tianbo@163.com

  • 基金资助:
    国家科技支撑计划项目(2013BAC03B04);国家自然科学基金项目(41301468)

Data Processing and Landslide Information Extraction Based on UAV Remote Sensing

CHEN Tianbo(), HU Zhuowei(), WEI Lai, HU Shunqiang   

  1. 1. College of Resource Environment and Tourism, Capital Normal University, Beijing 100048, China;
    2. Key Lab of Resources Environment and GIS, Capital Normal University, Beijing 100048, China;
    3. Key Lab of 3D Information Acquisition and Application, Ministry of Education, Capital Normal University, Beijing 100048, China
  • Received:2016-01-21 Revised:2016-06-23 Online:2017-05-20 Published:2017-05-27

摘要:

高分辨率的DEM和DOM数据是对地形地貌信息的准确描述,也是滑坡信息提取的重要数据源。首先,针对滑坡信息提取的要求,本文采用无人搭载微型单反相机的影像获取平台,结合野外测量的GPS数据,弥补了无人机POS信息精度低的劣势;针对无人机影像的特点,运用摄影测量基本原理与计算机视觉算法,获取高精度、高分辨率的DEM与DOM影像,保留了丰富的光谱与纹理信息。其次,借助ESP辅助工具获取了DOM影像的最佳分割尺度,并结合研究区地物特征构建了基于模糊分类与SVM算法相结合的决策树,运用面向对象的分类方法实现了对研究区内植被、道路、疑似滑坡区域的信息提取。最后,依照研究区地物分布的空间特征确定了高风险等级区域,并对该区域进行滑坡的形态与纹理分析以及精度评价,其中提取的疑似滑坡区域用户精度为91.44%、生产者精度为84.65%,结果表明无人机遥感在滑坡信息提取领域具有较高的应用价值。

关键词: 无人机遥感, 滑坡信息提取, 面向对象分类, 影像处理, DEM, DOM

Abstract:

The high-resolution DEM and DOM data is an accurate description of the topography and geomorphology, and it is also an important source data for landslide information extraction. At first,according to the requirement of landslide information extraction,we use the UAV platform equipped with mini SLR camera combined with the GPS data measured in the field, as the image acquisition method. According to the characteristics of the UAV images, we use the basic principle of photography measurement and computer vision algorithms to obtain the high-resolution DEM and DOM images, which greatly preserves the rich spectral and texture information. Then, with the help of the ESP auxiliary tool we get optimal segmentation scale of the DOM. Based on the fuzzy classification and SVM algorithm to construct a decision tree, which we used to achieve the object oriented classification and information extraction. Finally, according to the spatial feature and distribution of study area we determine the high risk area. By the morphology and texture analysis and accuracy assessment of the landslide area, we show that the producer’s accuracy and user's accuracy of the landslide area are 84.65% , 91.44%. The result proves that the UAV remote sensing has a high value in the field of landslide information extraction.

Key words: UAV remote sensing, landslide information extraction, object-oriented classification, image processing, DEM, DOM