Ga. Carpenter et al., ART NEURAL NETWORKS FOR REMOTE-SENSING - VEGETATION CLASSIFICATION FROM LANDSAT TM AND TERRAIN DATA, IEEE transactions on geoscience and remote sensing, 35(2), 1997, pp. 308-325
A new methodology for automatic mapping from Landsat thematic mapper (
TM) and terrain data, based on the fuzzy ARTMAP neural network, is dev
eloped. System capabilities are tested on a challenging remote sensing
classification problem, using spectral and terrain features for veget
ation classification in the Cleveland National Forest. After training
at the pixel level, system performance is tested at the stand level, u
sing sites not seen during training, Results are compared to those of
maximum likelihood classifiers, as well as back propagation neural net
works and K Nearest Neighbor algorithms, ARTMAP dynamics are fast, sta
ble, and scalable, overcoming common limitations of back propagation.
Best results are obtained using a hybrid system based on a convex comb
ination of fuzzy ARTMAP and maximum likelihood predictions, A prototyp
e remote sensing example introduces each aspect of data processing and
fuzzy ARTMAP classification. The example shows how the network automa
tically constructs a minimal number of recognition categories to meet
accuracy criteria, A voting strategy improves prediction and assigns c
onfidence estimates by training the system several times on different
orderings of an input set.