Share Email Print
cover

Proceedings Paper • new

Optimizing classification performance in an object-based very-high-resolution land use-land cover urban application
Author(s): Stefanos Georganos; Tais Grippa; Sabine Vanhuysse; Moritz Lennert; Michal Shimoni; Eléonore Wolff
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

This study evaluates the impact of three Feature Selection (FS) algorithms in an Object Based Image Analysis (OBIA) framework for Very-High-Resolution (VHR) Land Use-Land Cover (LULC) classification. The three selected FS algorithms, Correlation Based Selection (CFS), Mean Decrease in Accuracy (MDA) and Random Forest (RF) based Recursive Feature Elimination (RFE), were tested on Support Vector Machine (SVM), K-Nearest Neighbor, and Random Forest (RF) classifiers. The results demonstrate that the accuracy of SVM and KNN classifiers are the most sensitive to FS. The RF appeared to be more robust to high dimensionality, although a significant increase in accuracy was found by using the RFE method. In terms of classification accuracy, SVM performed the best using FS, followed by RF and KNN. Finally, only a small number of features is needed to achieve the highest performance using each classifier. This study emphasizes the benefits of rigorous FS for maximizing performance, as well as for minimizing model complexity and interpretation.

Paper Details

Date Published: 4 October 2017
PDF: 9 pages
Proc. SPIE 10431, Remote Sensing Technologies and Applications in Urban Environments II, 104310I (4 October 2017); doi: 10.1117/12.2278482
Show Author Affiliations
Stefanos Georganos, Univ. Libre de Bruxelles (Belgium)
Tais Grippa, Univ. Libre de Bruxelles (Belgium)
Sabine Vanhuysse, Univ. Libre de Bruxelles (Belgium)
Moritz Lennert, Univ. Libre de Bruxelles (Belgium)
Michal Shimoni, Royal Military Academy (Belgium)
Eléonore Wolff, Univ. Libre de Bruxelles (Belgium)


Published in SPIE Proceedings Vol. 10431:
Remote Sensing Technologies and Applications in Urban Environments II
Thilo Erbertseder; Nektarios Chrysoulakis; Ying Zhang; Wieke Heldens, Editor(s)

© SPIE. Terms of Use
Back to Top