Quantitatively Assessing Roads Extracted from High-Resolution Imagery

High spatial resolution images can be used for the extraction of road networks, however there are certain limits and properties of the resulting maps. This article focuses on the problem of quality when extracting roads from such data. The authors first discuss the possible methods for a road extraction process from remotely sensed images. They then propose a method to extract reference objects for the situations where the images are the only information available. Some criteria are then proposed for a quantitative assessment of the quality of extracted urban road networks. In the last section of the article, the method and criteria that the authors proposed are applied to a high resolution Ikonos image using a semi-automatic algorithm. In this case study, eleven photo-interpreters performed the extraction of one road. The reference toad and its tolerance zone were defined before the algorithm of road extraction was applied. The authors conclude that the proposed method for determining a reference object and its tolerance zone is not restricted to roads and should be easily adapted to other objects, such as buildings or rivers.

  • Availability:
  • Authors:
    • Péteri, Renaud
    • Couloigner, Isabelle
    • Ranchin, Thierry
  • Publication Date: 2004-12


  • English

Media Info

  • Media Type: Print
  • Features: Figures; Photos; References; Tables;
  • Pagination: pp 1449-1456
  • Serial:

Subject/Index Terms

Filing Info

  • Accession Number: 01002041
  • Record Type: Publication
  • Files: TRIS
  • Created Date: Jul 18 2005 10:43AM