Skip to main content
Conference paper, 2003

Remote Sensing Classification of Environmental Indicators

Ranneby Bo, Yu Jun


In Sweden 15 national environmental quality objectives have been defined together with specific environmental indicators to be analyzed repeatedly to follow relevant environmental trends. Remote sensing offers potential to assess wall-to-wall changes in the ecosystems. With the environmental indicators related to forestry as background the forest are divided into different classes. Remote sensing classification of this type of objects is complicated and reported classification rates are usually poor. To overcome most of the problems with traditional methods we use a new approach, where multispectral and change detection techniques are combined. Measures for quality assessment of the classified image are developed and applied. Furthermore, the importance of unbiased and high-precision estimates of the confusion matrix is emphasized. Even if the classified image gives incorrect proportions of the different classes unbiased area estimates can be derived from the confusion matrix


Remote sensing; classification; k-NN; probabilistic classifier; wavelet transform; information theory; quality assessment

Published in

Proceedings of ISRSE
2003, volume: Honolulu 2003
Publisher: ISRSE


30th International Symposium on Remote Sensing of Environment - Information for Risk Management and Sustainable Development

Authors' information

Yu, Jun
Swedish University of Agricultural Sciences, Department of Forest Economics
Swedish University of Agricultural Sciences, Department of Forest Economics

UKÄ Subject classification

Environmental Sciences related to Agriculture and Land-use
Forest Science

URI (permanent link to this page)