Automated Identification of Rivers and Shorelines in Aerial Imagery Using Image Texture

Abstract

A method has been developed which automatically extracts river and river bank locations from arbitrarily sourced high resolution (~lm) visual spectrum imagery without recourse to multi-spectral or even color information. This method relies on quantifying the difference in image texture between the relatively smooth surface of the river water and the rougher surface of the vegetated land or built environment bordering it and then segmenting the image into high and low roughness regions. The edges of the low roughness regions then define the river banks. The method can be coded in any language without recourse to proprietary tools and requires minimal operator intervention. As this sort of imagery is increasingly being made freely available through such services as Google Earth or Worldwind this technique can be used to extract river features when more specialized imagery or software is not available.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jan 01, 2011
Accession Number
ADA547608

Entities

People

  • Cheryl A. Blain
  • Paul Mckay
  • Robert Linzell

Organizations

  • United States Naval Research Laboratory

Tags

Communities of Interest

  • Energy and Power Technologies
  • Space

DTIC Thesaurus Topics

  • Aerial Photography
  • Computer Vision
  • Electronic Mail
  • Environment
  • Frequency Domain
  • Geometry
  • High Resolution
  • Hyperspectral Imagery
  • Identification
  • Images
  • Language
  • Military Research
  • Photography
  • Roughness
  • Spectra
  • Two Dimensional
  • Water

Readers

  • Agent-Based Social Robotics and Mobile-Assisted Learning in Virtual Environments.
  • Archaeological Resource Survey
  • Computer Vision.