[-] Show simple item record

dc.contributor.advisorJi, Wei, 1955-eng
dc.contributor.authorXu, Xiaofan, 1988-eng
dc.coverage.spatialMissouri -- Kansas Cityeng
dc.date.issued2014-07-30eng
dc.date.submitted2014 Springeng
dc.descriptionTitle from PDF of title page, viewed on July 30, 2014eng
dc.descriptionThesis advisor: Wei Jieng
dc.descriptionVitaeng
dc.descriptionIncludes bibliographical references (pages 85-93)eng
dc.descriptionThesis (M. S.)--Dept. of Geosciences. University of Missouri--Kansas City, 2014eng
dc.description.abstractIt has been a technical challenge to accurately detect urban wetlands with remotely sensed data by means of pixel-based image classification. This is mainly caused by inadequate spatial resolutions of satellite imagery, spectral similarities between urban wetlands and adjacent land covers, and the spatial complexity of wetlands in human-transformed, heterogeneous urban landscapes. Knowledge-based classification, with great potential to overcome or reduce these technical impediments, has been applied to various image classifications focusing on urban land use/land cover and forest wetlands, but rarely to mapping the wetlands in urban landscapes. This study aims to improve the mapping accuracy of urban wetlands by integrating the pixel-based classification with the knowledge-based approach. The study area is the metropolitan area of Kansas City, USA. SPOT satellite images of 1992, 2008, and 2010 were classified into four classes -- wetland, farmland, built-up land, and forestland -- using the pixel-based supervised maximum likelihood classification method. The products of supervised classification are used as the comparative base maps. For our new classification approach, a knowledge base is developed to improve urban wetland detection, which includes a set of decision rules of identifying wetland cover in relation to its elevation, spatial adjacencies, habitat conditions, hydro-geomorphological characteristics, and relevant geostatistics. Using ERDAS Imagine software's knowledge classifier tool, the decision rules are applied to the base maps in order to identify wetlands that are not able to be detected based on the pixel-based classification. The results suggest that the knowledge-based image classification approach can enhance the urban wetland detection capabilities and classification accuracies with remotely sensed satellite imageryeng
dc.description.tableofcontentsAbstract -- List of illustrations -- List of tables -- Acknowledgements -- Introduction -- Literature review -- Methodology -- Findings and analysis -- Discussion and conclusion -- Reference listeng
dc.format.extentxi, 94 pageseng
dc.identifier.urihttp://hdl.handle.net/10355/43573eng
dc.subject.lcshWetland managementeng
dc.subject.lcshWetlands -- Missouri -- Kansas Cityeng
dc.subject.lcshGeographic information systemseng
dc.subject.otherThesis -- University of Missouri--Kansas City -- Geoscienceseng
dc.titleA Knowledge-based approach of satellite image classification for urban wetland detectioneng
dc.typeThesiseng
thesis.degree.disciplineGeosciences (UMKC)eng
thesis.degree.grantorUniversity of Missouri-Kansas Cityeng
thesis.degree.levelMasterseng
thesis.degree.nameM. S.eng


Files in this item

[PDF]

This item appears in the following Collection(s)

[-] Show simple item record