Supporting the update of maps by object-oriented classification of orthophotos

摘要

The update of land use databases, if done on the basis of very high-resolution orthophotos, is usually carried out by means of visual interpretation, a both time-consuming and expensive task. In this paper an object-oriented classifier is used to support the update process by highlighting where changes have taken place. Basic land cover types are first identified in the orthophotos and then compared to the land use data to create a change map. The procedure was developed using true-color orthophotos and highly detailed land use data for a study area in Austria. Special emphasis was put on the identification of urban features. The resulting change map was compared to an independently updated version of the land use data base, showing the reliability of the results. The methodology was then tested on data sets provided by mapping agencies of Germany and Switzerland during the project "Change Detection" carried out within the framework of EuroSDR. The results are very consistent and even though they may not always meet the specific demands of the different mapping agencies, its potential for routine updating could be shown.