Improvement of classification accuracy by two neural networks and its application to land cover mapping

Author

Murai, Hiroshi ; Omatu, Sigeru ; Oe, Shunichiro

Author_Institution

Shikoku Univ., Tokushima, Japan

Volume

2

fYear

2000

fDate

2000

Firstpage

685

Abstract

In the recent works, the authors have proposed a hybrid system using a Kohonen´s self-organization feature mapping preprocessor (SOM) and a multi-layered neural network processor (BPM) to analyze remotely sensed data, and demonstrated the significance of outputs of the SOM preprocessor by a principal component analysis (PCA). In the present paper the authors apply the proposed system on three images taken by three optical sensors, namely, LANDSAT-TM, JERS1-OPS and SPOT2-HRV, to investigate the applicability of the system

Keywords

feedforward neural nets; geophysical signal processing; geophysical techniques; geophysics computing; image classification; principal component analysis; self-organising feature maps; terrain mapping; JERS1-OPS; Kohonen´s self-organization feature mapping preprocessor; LANDSAT-TM; SPOT2-HRV; accuracy; feedforward neural net; geophysical measurement technique; image classification; image processing; land cover mapping; land surface; multi-layered neural network processor; neural net; neural network; optical imaging; principal component analysis; remote sensing; terrain mapping; Cities and towns; Data mining; Data preprocessing; Multi-layer neural network; Neural networks; Pattern classification; Principal component analysis; Remote sensing; Satellites; Training data;

fLanguage

English

Publisher

ieee

Conference_Titel

Geoscience and Remote Sensing Symposium, 2000. Proceedings. IGARSS 2000. IEEE 2000 International

Conference_Location

Honolulu, HI

Print_ISBN

0-7803-6359-0

Type

conf

DOI

10.1109/IGARSS.2000.861671

Filename

861671