Neural network classification of multi-date satellite imagery
Experiments have been conducted on the classification of two-date SPOT imagery using neural networks. The basic neural network model which has been implemented is the multilayer perceptron net trained by the backpropagation algorithm. Both single networks, with about 100 nodes in total, and also hierarchical structures of smaller nets, have been developed. The classification accuracies achieved with these nets are in the range 80-90% for 20 land cover classes. Results are presented for experiments conducted in the Ardèche region of southern France.
Bibliographic Reference: Paper presented: IGARSS '91, Espoo (FI), June 3-6, 1991
Availability: Available from (1) as Paper EN 35994 ORA
Record Number: 199110433 / Last updated on: 1994-12-02
Original language: en
Available languages: en