Methodische Erweiterung der Landnutzungsklassifikation und Implementierung auf einem Transputernetzwerk

Abstract

The aim of the presented work is a comprehensive view of land use classification methods applying remote sensing data. Besides the standard approach of multispectral classification the theory of texture analysis is described and demonstrated by means of concrete examples. Next a variety of postclassification algorithms, which aggregate land cover types to land use classes, is discussed. In this context a postclassification method is introduced, that is based on a hierarchical rule system. The second part demonstrates the implementation of the algorithms on a transputer network. It focusses on the design of program shells, that allow a simple implementation of local image processing algorithms. Besides aspects of visualization and data compression are discussed. The application of these methods by mean of a land use classification for the design of mobile radio systems makes up the third part. An integrative classification approach including multispectral, textural and postclassification methods, is applied to a hybrid data set. Furthermore visual interpretations of orthophotos are performed as an alternative method. Comparison of both methods and evaluation based on reference data conclude the study.