One of the main goals of the Bavarian forest administration is the maintenance and reinforcement of the protective role of mountainous forests in the alpine region. For this purpose, there is a need of inventory and monitoring on a regular time interval, to obtain the necessary information about the status quo and possible dangerous conditions and developments among the forest stands. The terrain conditions and low accessibility of the region makes the collection of important forest stand parameters a costly affair. Remote Sensing and in particular the availability of newer high-resolution satellite images, offer a solution for this problem. However, the increasing spatial resolution of modern (satellite) sensors as well as the orographical terrain conditions make it necessary, to follow new ways of image classification, using the synergy effects derived from GIS information. In this study, the integration of the two domains, remote sensing and GIS, were analysed focussing on the needs of practical forest applications, using an object oriented and knowledge based image analysis system. Special interest was put on the verification of the geometric as well as the thematic accuracy of the data integrated in the GIS, because the combination and integration of high ground resolution remote sensing data in GIS analysis require the exact positioning and alignment of geographical data. Existing thematic data layers were used first to improve classification accuracy of remotely sensed data and then applied in updating the existing GIS data bases with the results of remote sensing data analysis. With the help of the different thematic GIS data layers, a knowledge based classification support was built up to optimise and also to validate the results from the classification of the remote-sensing-data. The well investigated interrelations and dependencies of the single factors, such as vegetation, site and landuse, are included in the evaluation of an allocation assistent based on a fuzzy-logic-rule-set. The use of auxiliary information could reduce the problem of the ambiguity of spectral signatures, which do not permit clearly a sufficiently exact separation of single vegetation classes. With the help of the knowledge- and object-based procedure an enhancement of the class choice in relation to a classification based on purely spectral information and also an updating or enlargement of existing databases was achieved.
«
One of the main goals of the Bavarian forest administration is the maintenance and reinforcement of the protective role of mountainous forests in the alpine region. For this purpose, there is a need of inventory and monitoring on a regular time interval, to obtain the necessary information about the status quo and possible dangerous conditions and developments among the forest stands. The terrain conditions and low accessibility of the region makes the collection of important forest stand parame...
»