Application of Geostatistical Methods and Wavelets to the Analysis of Hyperspectral Imagery and the Testing of a Moving Variogram

Abstract

The work in this report focuses on an analysis of the National Soil Inventory of England and Wales. The aim was to compare geostatistical methods, mainly ordinary kriging and factorial kriging and wavelet analysis, on a different kind of data from imagery. The data were from sampling locations on a 5-km grid. To provide an area as close to a square as possible for the wavelet analysis, just over 3000 points were selected from the total of over 5000. Two variables were selected for analysis, pH and zinc. The variogram: of pH showed that there was long-range trend in the data which meant that this had to be removed for the geostatistical analysis. Trend makes the geostatistical analysis more complex, whereas the wavelet analysis is not affected by it. Zinc was markedly skewed and the data were transformed to common logarithms for the geostatistical analysis, which again was not necessary for the wavelet analysis. The results have shown some interesting features. There appears to be no local non-stationarity in these data, which meant that kriging performed better than the wavelet analysis in terms of the distribution of the errors for the 10-km subsample. However, for the 40-km subsample the wavelet analysis performed better. The variograms for both properties were nested and the short-range variation was evident in the high frequency wavelet transform for the 20-km grid. The variogram can provide a guide as to what sampling interval should be focussed on in a multiresolution analysis using wavelets.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Nov 27, 2000
Accession Number
ADA389558

Entities

People

  • Margaret A. Oliver

Organizations

  • University of Reading

Tags

DTIC Thesaurus Topics

  • Agriculture
  • Coefficients
  • Data Sets
  • Ecology
  • Elements
  • Errors
  • Frequency
  • Grids
  • Hyperspectral Imagery
  • Inventory
  • Normal Distribution
  • Soil Surveys
  • Statistics
  • Surveys
  • United States
  • Urban Areas
  • Wavelet Transforms

Readers

  • Image Processing and Computer Vision.
  • Regression Analysis.