电子学报 ›› 2019, Vol. 47 ›› Issue (2): 489-494.DOI: 10.3969/j.issn.0372-2112.2019.02.033

• 学术论文 • 上一篇    下一篇

模糊区域级MRF方法在城镇自动识别中的应用

陈荣元, 徐雪松, 申立智, 刘跃华, 陈浪   

  1. 湖南商学院湖南省移动电子商务协同创新中心, 移动商务智能湖南省重点实验室, 新零售虚拟现实技术湖南省重点实验室, 湖南长沙 410205
  • 收稿日期:2017-10-30 修回日期:2018-05-22 出版日期:2019-02-25 发布日期:2019-02-25
  • 通讯作者: 徐雪松
  • 作者简介:陈荣元 男,1976年6月生,江苏兴化人,2010年于武汉大学获得博士学位,计算机学会高级会员,现为湖南商学院教授,主要研究方向为图像处理和人工智能.E-mail:chenrongyuan@126.com;申立智 男,1983年6月生,湖南邵东人,2011年于长沙理工大学获得硕士学位,主要研究方向为图像处理和数据挖掘.E-mail:slzzx1983@163.com
  • 基金资助:
    国家自然科学基金(No.41101425,No.61471170);教育部-中国移动科研基金(No.MCM20170506);湖南省教育厅资助科研项目(No.16A114,No.17B145);湖南省重点研发计划项目(No2018GK2058);湖南省自然科学基金(No.2016JJ2070,No.2017JJ3132)

Fuzzy Regional MRF Method for Automatic Identification of Urban

CHEN Rong-yuan, XU Xue-song, SHEN Li-zhi, LIU Yue-hua, CHEN Lang   

  1. Mobile E-business Collaborative Innovation Center of Hunan Province, Key Laboratory of Hunan Province for Mobile Business Intelligence, Key Laboratory of Hunan Province for New Retail Virtual Reality Technology, Hunan University of Commerce, Changsha, Hunan 410205, China
  • Received:2017-10-30 Revised:2018-05-22 Online:2019-02-25 Published:2019-02-25

摘要: 针对遥感影像中城镇区域内外的自然地物难以区分,城镇区域不易完整识别的问题,提出一种对象级模糊MRF识别方法.该方法首先通过光谱信息和空间梯度分析得到城镇种子点(人造地物顶部点和阴影点);然后由均值漂移算法过分割影像;再对过分割区域建立MRF,在迭代过程中用MRF的条件概率矩阵代替模糊C均值聚类算法的隶属度矩阵,并保持包含种子点的区域类别不变,从而实现城镇识别.对于QuickBird和Ikonos遥感影像,该模型能够兼顾城镇区域自动识别过程中的随机性与模糊性,很好地利用了空间相关信息,有效识别出了城镇区域.

关键词: 城镇识别, 模糊C均值算法, 空间信息, 马尔可夫模型

Abstract: In order to effectively distinguish the natural objects inside the town from outside objects,and completely identify the urban regions in remote sensing image,a fuzzy geographic object-based MRF method is proposed.Firstly,the seed points of the town,i.e.,the top and shadow points of artificial ground objects,are firstly obtained by analyzing spectral information and spatial gradient.Then the over-segmented regions of the origin image are obtained by Mean Shift algorithm.Finally,a MRF is established over regions.and the membership matrix in the fuzzy C-means clustering algorithm is replaced by the conditional probability matrix in the MRF in an iterative manner.Meanwhile the categories of the regions containing the seed points are kept unchanged.For QuickBird and Ikonos remote sensing images,the proposed model can simultaneously deal with both the stochastic and fuzzy nature of images,and effectively intergate the space information,and thus benefit the identification of town.

Key words: urban identification, fuzzy C-means algorithm, spatial information, Markov random field model

中图分类号: