Information Fusion for Urban Road Extraction From VHR Optical Satellite Images

Zelang Miao, Wenzhong Shi, Alim Samat, Gianni Lisini, Paolo Gamba

Research output: Journal article publicationJournal articleAcademic researchpeer-review

20 Citations (Scopus)

Abstract

This paper presents a novel method exploiting fusion at the information level for urban road extraction from very high resolution (VHR) optical satellite images. Given a satellite image, we explore spectral and shape features computed at the pixel level, and use them to select road segments using two different methods (i.e., expectation maximization clustering and linearness filtering). A road centerline extraction method, which is relying on the outlier robust regression, is subsequently applied to extract accurate centerlines from road segments. After that, three different sets of information fusion rules are applied to jointly exploit results from these methods, which offer ways to address their own limitations. Two VHR optical satellite images are used to validate the proposed method. Quantitative results prove that information fusion following centerline extraction by multiple techniques is able to produce the best accuracy values for automatic urban road extraction from VHR optical satellite images.
Original languageEnglish
Article number7337372
Pages (from-to)1817-1829
Number of pages13
JournalIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Volume9
Issue number5
DOIs
Publication statusPublished - 1 May 2016

Keywords

  • Centerline
  • expectation maximization (EM)
  • information fusion
  • linearness filter
  • RANdom SAmple Consensus (RANSAC)

ASJC Scopus subject areas

  • Computers in Earth Sciences
  • Atmospheric Science

Cite this