Community Research and Development Information Service - CORDIS

Abstract

The data collected by satellite sensors are spatially integrated over a pixel size of 20 - 30 m. A pixel of this size often contains a variety of small scale features and vegetation types, making it difficult to put pixels into meaningful land cover classes. The multilayer perceptron (MLP) model is superior to statistically based classifers in this repect. This paper shows that MLP networks can usefully encode feature vectors in their last hidden layer. These can be used as input to an LVQ network. The potential use of hybrid MLP/LVQ networks to improve performance in a real-world complex classification task is thus demonstrated.

Additional information

Authors: HERNANDEZ R, JRC Ispra (IT);VARFIS A, JRC Ispra (IT);KANELLOPOULOS I, JRC Ispra (IT);WILKINSON G, JRC Ispra (IT)
Bibliographic Reference: Paper presented: International Conference on Artificial Neural Networks, Brighton (GB), Sept. 4-7, 1992
Availability: Available from (1) as Paper EN 36803 ORA
Record Number: 199210708 / Last updated on: 1994-12-02
Category: PUBLICATION
Original language: en
Available languages: en