TY - JOUR
T1 - Supervised classification of civil air patrol (CAP)
AU - Sava, Elena
AU - Clemente-Harding, Laura
AU - Cervone, Guido
N1 - Funding Information:
Work performed under this project has been partially funded by the Office of Naval Research (ONR) Award #N00014-14-1-0208 (PSU #171570). We wish to thank Drs. Andris and Brooks for their comments that helped improve the present manuscript.
Publisher Copyright:
© 2016, Springer Science+Business Media Dordrecht.
PY - 2017/3/1
Y1 - 2017/3/1
N2 - The mitigation and response to floods rely on accurate and timely flood assessment. Remote sensing technologies have become the de facto approach for observing the Earth and its environment. However, satellite remote sensing data are not always available, and it is crucial to develop new techniques to complement them with additional sources. This research proposes a new methodology based on machine learning algorithms to automatically identify water pixels in Civil Air Patrol (CAP) aerial imagery. Specifically, a wavelet transformation is paired with multiple classifiers to build models that discriminate water and non-water pixels. The learned classification models are first tested against a set of control cases and then used to automatically classify each image separately. Lastly, for each pixel in an image, a measure of uncertainty is computed as a proportion of the number of models that classify the pixel as water. The proposed methodology is tested on imagery collected during the 2013 Colorado flood.
AB - The mitigation and response to floods rely on accurate and timely flood assessment. Remote sensing technologies have become the de facto approach for observing the Earth and its environment. However, satellite remote sensing data are not always available, and it is crucial to develop new techniques to complement them with additional sources. This research proposes a new methodology based on machine learning algorithms to automatically identify water pixels in Civil Air Patrol (CAP) aerial imagery. Specifically, a wavelet transformation is paired with multiple classifiers to build models that discriminate water and non-water pixels. The learned classification models are first tested against a set of control cases and then used to automatically classify each image separately. Lastly, for each pixel in an image, a measure of uncertainty is computed as a proportion of the number of models that classify the pixel as water. The proposed methodology is tested on imagery collected during the 2013 Colorado flood.
UR - http://www.scopus.com/inward/record.url?scp=85001759394&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85001759394&partnerID=8YFLogxK
U2 - 10.1007/s11069-016-2704-3
DO - 10.1007/s11069-016-2704-3
M3 - Article
AN - SCOPUS:85001759394
VL - 86
SP - 535
EP - 556
JO - Natural Hazards
JF - Natural Hazards
SN - 0921-030X
IS - 2
ER -