An Interactive Model for Atmospheric Correction in Satellite Images,

Abstract

The presence of the terrestrial atmosphere is a constant problem in satellite image classification. This paper describes a procedure to test quantitatively the performance of an atmospheric correction algorithm. This specific algorithm was developed at INPE, based on O'Neil's model. It is difficult to obtain a reasonable atmospheric model for Brazilian conditions, due to lack of aerosol data. The procedure consists of correcting two images of the desired scene, for different days, changing the aerosol optical depth, and comparing at selected areas, the contrast and the correlation between the images. For a better correction, the contrast improves, as well as the correlation. Based on the previous information, it is possible to select the proper aerosol optical depth. Care should be taken, however, since an exaggerated contrast improvement by itself or a good correlation alone are not sufficient. Using both criteria, a better judegement can be made. Scenes that have undergone considerable temporal changes, as those at different stages of a given crop, should be avoided. In order to make the interactive model practical, it was necessary to calculate quickly and interactively the atmospheric transmittance. This was done by means of a modified LOWTRAN program.

Document Details

Document Type
Technical Report
Publication Date
Jun 01, 1982
Accession Number
ADP002075

Entities

People

  • A. E. C. Pereira
  • L. A. V. Dias
  • N. L. Vijaykumar

Tags

Communities of Interest

  • Space

DTIC Thesaurus Topics

  • Algorithms
  • Argentina
  • Artificial Satellites
  • Atmospheres
  • Classification
  • Contrast
  • Environment
  • Image Classification
  • Remote Sensing
  • Transmittance

Fields of Study

  • Environmental science

Readers

  • Computer Vision.
  • Neural Network Machine Learning.
  • Spectroscopy.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference
  • AI & ML - Neural Networks
  • Space
  • Space - Space Objects