测绘通报 ›› 2019, Vol. 0 ›› Issue (6): 11-15.doi: 10.13474/j.cnki.11-2246.2019.0175

• 学术研究 • 上一篇    下一篇

改进分水岭方法下的遥感图像水体提取

冯朝晖1, 李芹1, 韩留生2, 薛国超3, 赵红颖1   

  1. 1. 北京大学地球与空间科学学院, 北京 100871;
    2. 山东理工大学建筑工程学院, 山东 淄博 255000;
    3. 山东科技大学测绘科学与工程学院, 山东 青岛 266590
  • 收稿日期:2018-10-17 出版日期:2019-06-25 发布日期:2019-07-01
  • 通讯作者: 赵红颖。E-mail:zhaohy@pku.edu.cn E-mail:zhaohy@pku.edu.cn
  • 作者简介:冯朝晖(1994-),女,硕士生,主要研究方向为遥感数字图像处理及无人机组网任务规划。E-mail:912735720@qq.com
  • 基金资助:

    国家重点研发计划(2017YFB0503003)

Remote sensing image water body extraction based on improved watershed

FENG Zhaohui1, LI Qin1, HAN Liusheng2, XUE Guochao3, ZHAO Hongying1   

  1. 1. School of Earth and Space Sciences, Peking University, Beijing 100871, China;
    2. School of Civil and Architectural Engineering, Shandong University of Technology, Zibo 255000, China;
    3. College of Surveying and Mapping Science and Engineering, Shandong University of Science and Technology, Qingdao 266590, China
  • Received:2018-10-17 Online:2019-06-25 Published:2019-07-01

摘要:

针对当下全球的水域环境监测中水体提取不完整、与周围地物区分错误等问题,以山东潍河为研究对象,提出了一种改进分水岭分割方法。首先提取目标水体的光谱信息分量作为前景图,然后将膨胀后的水体对象作为背景图,利用前景图与背景图叠加生成标记图,最后利用标记图通过分水岭变换对原图像进行分割以实现水体信息的自动提取。本文将试验结果与OpenCV中手动标记种子点的分水岭算法、基于Canny边缘的分水岭算法、结合形态滤波和标记的分水岭分割方法结果进行了对比。结果表明,本文算法比用种子点手动标记的方法更加自动化,提取的水体更加完整准确;相比于Canny边缘方法又避免了过度分割;比结合形态滤波标记的方法也更加完整准确,目标水体明显,对于水体提取是一种自动化的有效方法。

关键词: 水体提取, 光谱信息分量, 形态学膨胀, 分水岭变换

Abstract:

In view of the incomplete and the misidentification of surrounding objects of water extraction in the current global water environment monitoring, this paper takes the Shandong Weihe River as the research object, and the improved watershed segmentation method designed by this paper is adopted. Firstly, the spectral information component of the target water body is extracted as the foreground image, then use the expanded water object as the background image. And the foreground image and the background image are superimposed to generate a marker image. The marker image is used by the watershed transformation to realize the automatic extraction of the water body information. The results of this paper are compared with the watershed algorithm of manual marker seed point in OpenCV, the watershed algorithm based on Canny edge, the watershed segmentation method combined with morphological filtering and labeling. The experimental results show that the algorithm designed in this paper is more automated than the seed point watershed algorithm in OpenCV. And the extracted water body is more complete and accurate. The method in this paper avoids the over-segmentation compared to the Canny edge method. Compared with the method of combining morphological filtering markers, the method is more complete and accurate in extracting the target water body. In conclusion, this paper's method is an automatic and effective method for water body extraction.

Key words: water body extraction, spectral information component, morphological expansion, watershed transformation

中图分类号: