Automated detection of structures from remote sensing data
Liimatainen, Kaisa Maria (2014)
2014
Tietotekniikan koulutusohjelma
Tieto- ja sähkötekniikan tiedekunta - Faculty of Computing and Electrical Engineering
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2014-12-03
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tty-201412051587
https://urn.fi/URN:NBN:fi:tty-201412051587
Tiivistelmä
An objective method for determining the natural state of a mire is needed to preserve mire biodiversity in Finland. Ditches and roads are important descriptors of the natural state, so the objective of this thesis is to detect these structures in mire surroundings.
A digital terrain model created from LiDAR data is used for ditch classification and orthophotographs and various LiDAR data are used to detect roads. The classification is done with logistic regression classifier which selects best features for classification from a large feature set. In ditch detection polynomial modeling is used to connect broken segments.
Artificial drainage networks were detected well with the method and polynomial modeling improved the results. The percentage of found ditch points from all ditch points was 90.51 before polynomial modeling and 97.27 after. The road detection accuracy did not correspond to values obtained from ditch detection. Yet the ditch detection results indicate that logistic regression classification is a suitable method for this application. For successful classification the feature set needs to be large enough and the training set has to be comprehensive.
Artificial drainage network information will later be used in determining the extent of mire drainage and modeling waterflow patterns.
A digital terrain model created from LiDAR data is used for ditch classification and orthophotographs and various LiDAR data are used to detect roads. The classification is done with logistic regression classifier which selects best features for classification from a large feature set. In ditch detection polynomial modeling is used to connect broken segments.
Artificial drainage networks were detected well with the method and polynomial modeling improved the results. The percentage of found ditch points from all ditch points was 90.51 before polynomial modeling and 97.27 after. The road detection accuracy did not correspond to values obtained from ditch detection. Yet the ditch detection results indicate that logistic regression classification is a suitable method for this application. For successful classification the feature set needs to be large enough and the training set has to be comprehensive.
Artificial drainage network information will later be used in determining the extent of mire drainage and modeling waterflow patterns.