A comparison of neural network and expert system methods for analysis of remotely-sensed imageryFunded under: JRC-REMSENS 4C
This paper describes an experimental comparison which has been made between two alternative methods of image classification; one based on a neural network and the other on a rule-based expert system. Both methods were applied to the same image data. The results show that both methods give useful performance improvements in comparison with more traditional parametric classifiers. It was also found that the performance level attained by the two approaches was approximately the same. The neural network was, however, faster to develop although the expert system was much more transparent and easier for a user to understand.
Bibliographic Reference: Paper presented: Geoscience and Remote Sensing Symposium (INGARSS'92), Houston, Texas (US), May 26-29, 1992
Availability: Available from (1) as Paper EN 36932 ORA
Record Number: 199210948 / Last updated on: 1994-12-02
Original language: en
Available languages: en