Optimization Strategy of a Stacked Autoencoder and Deep Belief Network in a Hyperspectral Remote-Sensing Image Classification Model

Table 5

Confusion matrix of DBN-SR classification accuracy for HySpex data.

	Water	Vegetation	Concrete road	Magma rock	Steel plate	Glass	Wall	Total	Accuracy (%)
Water	379	9	2	0	0	0	0	390	97.18
Vegetation	15	368	0	0	0	0	0	383	96.08
Concrete road	7	41	401	0	0	0	0	449	89.31
Magma rock	0	0	0	302	16	0	38	356	84.83
Steel plate	0	0	0	0	337	8	32	377	89.39
Glass	0	0	0	0	0	194	17	211	91.94
Wall	0	0	0	7	5	5	394	411	95.86
Total	401	418	403	309	358	207	481

Overall accuracy = 2375/2577 = 92.16%; kappa coefficient = (0.9216 − 0.1481)/(1 − 0.1481) = 0.91.