Nonparametric classification and probabilistic classifiers with environmental and remote sensing applications
2007 (English)In: Asymptotic Theory in Probability and Statistics with Applications / [ed] Tze Leung Lai, Qi-Man Shao, Lianfen Qian, Beijing: Higher Education Press, 2007, 388-436 p.Chapter in book (Refereed)
National and international policies today require environmental monitoring and follow-up systems that detect, in a quality assured way, changes over time in land use and landscape indicators. Remote sensing of satellite images offers great potential to assess wall-to-wall changes in the health of ecosystems and identify risks. Questions related to environmental health and spatial patterns call for new statistical tools. We present in this chapter some new developments on the classification of land use and spatial indicators using multispectral and multitemporal satellite images. They are developed under non-standard conditions - conditions under which many statistical methods do not work properly but frequently appear in environmental and remote sensing applications. Error rates of traditional remote sensing classification methods are usually quite high but can be improved by (1) denoising the images using the wavelet transform, (2) reclassification using Markov random field approaches, or (3) applying new classification algorithms based on nonparametric classifiers. In order to assess quality of classification at pixel level, a new concept, the probabilistic classifiers, is introduced. These classifiers are useful for measuring uncertainty at pixel level and obtaining reliable area estimates locally. Results from both simulation studies and real applications are presented.
Place, publisher, year, edition, pages
Beijing: Higher Education Press, 2007. 388-436 p.
, Advanced Lectures in Mathematics, 2
Probability Theory and Statistics
Research subject Mathematical Statistics
IdentifiersURN: urn:nbn:se:umu:diva-63688ISBN: 978-7-04-022152-7OAI: oai:DiVA.org:umu-63688DiVA: diva2:582330