Title :
Evaluation methods for curvilinear feature extraction
Author :
Doucette, Peter ; Martin, Ann ; Kavanagh, Chris ; McIntyre, Tim ; Barton, Steven ; Grodecki, Jacek ; Malitz, Seth ; Tang, Matthew ; Nolting, Josh
Author_Institution :
Contractor for NGA, Reston, VA, USA
Abstract :
The application of quantitative performance evaluation methods can provide useful insights in determining the utility of computer-assisted methods for delineating geographic features from remotely sensed images. Evaluation concepts are demonstrated with road centerlines in particular, but are applicable to similar feature types such as paths, trails, or rivers. The two comparative measures used to differentiate conventional versus computer-assisted delineation are 1) user clock time, and 2) spatial consistency. Our evaluation results with road centerlines demonstrate how such quantitative analyses can be used to determine the utility of computer-assisted methods from both developmental and operational perspectives.
Keywords :
feature extraction; geographic information systems; performance evaluation; remote sensing; roads; computer-assisted delineation; curvilinear feature extraction; geographic feature delineation; quantitative performance evaluation methods; remotely sensed images; road centerlines; spatial consistency; user clock time; Application software; Cams; Clocks; Computer vision; Feature extraction; Geographic Information Systems; Pattern recognition; Rivers; Roads; Time measurement;
Conference_Titel :
Applied Imagery Pattern Recognition Workshop (AIPRW), 2009 IEEE
Conference_Location :
Washington, DC
Print_ISBN :
978-1-4244-5146-3
Electronic_ISBN :
1550-5219
DOI :
10.1109/AIPR.2009.5466322