测绘通报 ›› 2021, Vol. 0 ›› Issue (2): 49-53.doi: 10.13474/j.cnki.11-2246.2021.0042

• 学术研究 • 上一篇    下一篇

融合随机森林和超像素分割的建筑物自动提取

陈利燕, 林鸿, 吴健华   

  1. 广州市城市规划勘测设计研究院, 广东 广州 510060
  • 收稿日期:2020-03-04 出版日期:2021-02-25 发布日期:2021-03-09
  • 通讯作者: 林鸿。E-mail:linhong83762173@163.com
  • 作者简介:陈利燕(1981-),女,博士,高级工程师,研究方向为空间数据更新与融合。E-mail:546535531@qq.com
  • 基金资助:
    广州市博士后启动基金(201517040001);住建部科技计划(2017-KB-016);广州市科技计划(201702020201);广州市工信委信息化项目(GZIT2016-A5-147)

Building extraction based on random forest and superpixel segmentation

CHEN Liyan, LIN Hong, WU Jianhua   

  1. Guangzhou Urban Planning & Design Survey Research Institute, Guangzhou 510060, China
  • Received:2020-03-04 Online:2021-02-25 Published:2021-03-09

摘要: 建筑物是城市空间的重要部分,建筑物信息的提取对基础地理空间数据库更新、城市规划、城市动态监测等具有重要意义。基于遥感影像数据提取建筑物信息具有非常广泛的应用,本文提出了一种基于随机森林和超像素分割算法,并从机载激光点云和数字航空影像数据中自动提取建筑物的方法。试验选取广州市海珠区某处为研究区域,结果表明:在一般的城市区,90%以上建筑物可以准确快速提取,平均准确性和完整性均为90%左右,本文提出的方法具有良好的应用前景。

关键词: 建筑物提取, 高分辨率遥感影像, 激光雷达点云, 随机森林, 超像素分割

Abstract: Buildings are an important part of urban space. The extraction of building information is of great significance for basic geospatial database updates, urban planning, urban dynamic monitoring. Building information extraction based on remote sensing images data has a very wide range of applications. This study proposes a method based on random forest and super pixel segmentation algorithm, and automatically extracting buildings from airborne laser point cloud and digital aerial image data. The experiment selects a certain area in Haizhu district, Guangzhou as the research area. The results show that:In a general urban area, more than 90% of buildings can be extracted accurately and quickly, with an average accuracy and completeness of about 90%. The method proposed in this paper has a good application prospect.

Key words: building boundary extraction, high-resolution optical images, LiDAR, random forest, superpixel segmentation

中图分类号: