TY - GEN
T1 - Optimizing classification performance in an object-based very-high-resolution land use-land cover urban application
AU - Georganos, Stefanos
AU - Grippa, Tais
AU - Vanhuysse, Sabine
AU - Lennert, Moritz
AU - Shimoni, Michal
AU - Wolff, Eléonore
N1 - Publisher Copyright:
© 2017 SPIE.
PY - 2017
Y1 - 2017
N2 - 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.
AB - 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.
KW - feature selection
KW - object based image analysis
KW - random forest
KW - supervised classification
KW - support vector machine
UR - http://www.scopus.com/inward/record.url?scp=85040075014&partnerID=8YFLogxK
U2 - 10.1117/12.2278482
DO - 10.1117/12.2278482
M3 - Conference contribution
AN - SCOPUS:85040075014
T3 - Proceedings of SPIE - The International Society for Optical Engineering
BT - Remote Sensing Technologies and Applications in Urban Environments II
A2 - Erbertseder, Thilo
A2 - Zhang, Ying
A2 - Chrysoulakis, Nektarios
PB - Society of Photo-Optical Instrumentation Engineers
T2 - Remote Sensing Technologies and Applications in Urban Environments II 2017
Y2 - 11 September 2017 through 13 September 2017
ER -