Computer processing of multi-layer imagery data

Abstract

The technique and methodology of computer-aided image processing has become a valuable tool for handling, processing and analysing multi-spectral imagery data acquired from sensors in aircraft and satellites. This paper describes some of the background and the rationale for performing computer processing and also discusses types and kinds of operations which come into use. The main purpose of the paper is to give a description of a software system for performing various tasks in computer-aided image processing. One interesting part of this system is a supervised classification program system for pixelwise (or objectwise) multi-layer classification. The classification algorithm performs class assignment in a linearly transformed space. It is a classifier whose c.p.u. time is approximately linearly dependent upon the number of layers (e.g. spectral features) compared with an approximate quadratic dependence for a classical implementation of the maximum-likelihood Bayes classifier for normally-distributed features. Examples of pilot and application studies of various problems in environmental monitoring and land use are also given. Forest monitoring is considered to be an important application in Sweden. The paper also shows the importance of incorporating non-spectral topographic) data for increased classification accuracy of detecting clear-cut areas.