Comparing Supervised and Unsupervised Classifiers for Multispectral Image AnalysisProceedings of the Intelligent Engineering Systems through Artificial Neural Networks Conference, Volume 20 (2010)
This paper compares supervised and unsupervised classification of using satellite multi-spectral images for monitoring vegetation growth on lakes and its effect on water quality. The area of interest (AOI) is Lake Tyler which is the main water source for the City of Tyler. It is important to maintain the water quality by monitoring various parameters such as the amount of vegetation growth, surface area covered by water, water pollution, etc. Traditional field based mapping and monitoring present several challenges including inaccessibility and in identifying dynamic changes. Multi-spectral images from Landsat-5 Thematic Mapper (TM), along with the ground truth provided by the Texas Plants and Wildlife Department are used for the analysis. Thematic maps are generated using the maximum likelihood and isodata classifiers. The categories of interest include vegetation on Lake, water, vegetation on land and land. Classification results are evaluated and compared using measures such as the user's accuracy, producer's accuracy, overall accuracy, and Kappa coefficient.
- industrial plants,
- water pollution
Publication DateNovember, 2010
LocationSt. Louis, MO
Citation InformationArun D Kulkarni and Kiran Parimi. "Comparing Supervised and Unsupervised Classifiers for Multispectral Image Analysis" Proceedings of the Intelligent Engineering Systems through Artificial Neural Networks Conference, Volume 20 (2010)
Available at: http://works.bepress.com/arun-kulkarni/45/