测绘通报 ›› 2019, Vol. 0 ›› Issue (7): 69-72,126.doi: 10.13474/j.cnki.11-2246.2019.0221

• 技术交流 • 上一篇    下一篇

利用OpenStreetMap数据进行高空间分辨率遥感影像分类

郝怀旭, 万太礼, 罗年学   

  1. 武汉大学测绘学院, 湖北 武汉 430079
  • 收稿日期:2018-07-12 出版日期:2019-07-25 发布日期:2019-07-31
  • 作者简介:郝怀旭(1994-),男,硕士,研究方向为遥感图像分析、地理信息工程。E-mail:2012301610342@whu.edu.cn
  • 基金资助:
    国家重点研发计划(2017YFC1405300)

High-resolution remote sensing image classification using OpenStreetMap data

HAO Huaixu, WAN Taili, LUO Nianxue   

  1. School of Geodesy and Geomatics, Wuhan University, Wuhan 430079, China
  • Received:2018-07-12 Online:2019-07-25 Published:2019-07-31

摘要: 针对高分辨率遥感影像分类样本标注困难的问题,提出了一种利用OpenStreetMap (OSM)数据自动获取标注样本的方法。与现有的利用OSM数据进行分类的方法不同,该方法加入了空间特征以弥补单独使用光谱特征分类的不足。首先,基于OSM数据提供的地物类别和位置信息进行样本标注,为了降低OSM数据中少量错误信息对分类结果的影响,采用聚类分析的方法对样本进行提纯;其次,使用形态学轮廓来提取影像的结构特征,挖掘高分辨率遥感影像丰富的空间信息,与光谱特征相叠加并输入分类器进行分类。试验证明,本文提出的方法能够有效避免人工样本标注所需要的人力物力;同时,联合影像的光谱空间特征能够更好地描述地物特性,得到较高的分类精度。

关键词: 样本标注, OpenStreetMap, 形态学轮廓, 聚类分析, 高分辨率遥感影像

Abstract: To address the difficulties of labeling samples for high-resolution remote sensing images, an automatic method of labeling samples employing OpenStreetMap (OSM) data is proposed in this paper. Different from the existing approaches, the proposed method adopts spatial features to supplement the shortcomings of using only spectral features. Firstly, samples are labeled using the category and position information of OSM data. As OSM data may contain errors, cluster analysis is utilized to refine the derived samples. Besides, to exploit the abundant spatial information provided by high resolution remote sensing images, morphological profiles are used to describe the structural features of the images. The spatial features as well as spectral features are combined for classification. Experiments show that the proposed method can significantly avoid the manpower and material resources required for labeling samples artificially. Meanwhile, the derived samples and the spectral-spatial features both contribute to the classification accuracy.

Key words: sample labeling, OpenStreetMap, morphological profiles, cluster analysis, high-resolution remote sensing images

中图分类号: