Robust Statistics for Classification of Remote Sensing Data

Dyah E. Herwindiati (1) , Maman A. Djauhari (2) , Luan Jaupi (3)

1 UI - University of Indonesia
2 UTM - Universiti Teknologi Malaysia
3 CEDRIC - MSDMA - CEDRIC. Méthodes statistiques de data-mining et apprentissage

The classification of remote sensing data from Landsat 7 satellite is considered, and an area under investigation is Jakarta Province. The supervised land classification is done with two processes: the training sites and classification process. A robust computationally efficient approach is applied for training site to deal with the large remote sensing data set of Jakarta. The objective of this paper is to introduce the depth function for robust estimation of a multivariate location parameter minimizing vector variance for classification of green space at Jakarta Province.

Mots clés: Data depth; minimum vector variance; Mahalanobis distance; robust statistic; remote sensing; robust estimation

Mots clés: statistique robuste; la télédétection; variance minimale vector; Distance de Mahalanobis; Profondeur de données

Domaines: Statistiques [stat]

Dyah E. Herwindiati, Maman A. Djauhari, Luan Jaupi. Robust Statistics for Classification of Remote Sensing Data. 20th International Conference on Computational Statistics. COMPSTAT 2012, Aug 2012, Limassol, Cyprus. pp.317-328. ⟨hal-02468060⟩