Title :
Selection of feature variables in spatial discrimination of remotely-sensed satellite imagery
Author_Institution :
Fac. of Integrated Arts & Sci., Hiroshima Univ., Japan
Abstract :
Statistical discriminant procedures based on multi-spectral images are widely used for land-cover classification. However, all images available for discrimination are not always useful for discrimination. The author considers a spatially-correlated multivariate normal distribution for the multispectral data. Under the local continuity of land-cover categories, they propose a statistical techniques for finding the best and parsimonious subset of the multispectral feature variables
Keywords :
feature extraction; geophysical signal processing; geophysical techniques; image classification; multidimensional signal processing; remote sensing; terrain mapping; feature extraction; feature variables selection; geophysical measurement technique; image classification; image processing; land surface; land-cover classification; local continuity; multi-spectral image; multispectral image; remote sensing; satellite imagery; spatial discrimination; spatially-correlated multivariate normal distribution; statistical discriminant procedure; statistical discrimination; terrain mapping; Art; Covariance matrix; Gaussian distribution; Input variables; Multispectral imaging; Pixel; Remote sensing; Satellites; Vectors;
Conference_Titel :
Geoscience and Remote Sensing Symposium, 1999. IGARSS '99 Proceedings. IEEE 1999 International
Conference_Location :
Hamburg
Print_ISBN :
0-7803-5207-6
DOI :
10.1109/IGARSS.1999.772104
