Multi Angle Imaging With Spectral Remote Sensing for Scene Classification

Abstract

Scene classification is studied here using the tool of texture analysis of multi-angle high-spatial resolution panchromatic and multi-spectral imagery. This study analyses the BRDF (Bidirectional Reflectance Distribution Function) impact and effectiveness of texture analysis on terrain classification within Fresno County area in state of California. QuickBird panchromatic (0.61 meter) and multispectral (2.44 meter) imagery collected in July 2003 are examined to determine the impact of adding multi-angles and filtered texture information to the standard MSI classification approaches. Four images were collected, with view angles from -64 to +64 , including a nadir view. Texture filter function and maximum likelihood classifier are used in this study. Both texture analysis and the results of classifications using multi-angle (BRDF) information are promising. Fine discrimination of similar soil classes was produced by the BRDF variations in the high-spatial resolution panchromatic image. Texture analysis results depended on the directionality of the gray level co-occurrence matrix (GLCM) calculation. Combining the different modalities of analysis did not improve the overall classification, perhaps illustrating the consequences of the Hughes paradox (Hughes, 1968).

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Mar 01, 2005
Accession Number
ADA432973

Entities

People

  • Sunyaruk Prasert

Organizations

  • Naval Postgraduate School

Tags

Communities of Interest

  • Biomedical
  • Sensors
  • Space

DTIC Thesaurus Topics

  • Backscattering
  • California
  • Classification
  • Detectors
  • Distribution Functions
  • Electromagnetic Radiation
  • Forward Scattering
  • Geographic Regions
  • Geometry
  • Image Processing
  • Image Registration
  • Information Science
  • Machine Learning
  • Optical Properties
  • Reflectance
  • Remote Sensing
  • Urban Areas

Readers

  • Atmospheric Remote Sensing.
  • Computer Vision.