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

Table 3

Confusion matrix for the SAE-SR classification accuracy of HySpex data.

	Water	Vegetation	Concrete road	Magma rock	Steel plate	Glass	Wall	Total	Accuracy (%)
Water	377	8	5	0	0	0	0	390	96.67
Vegetation	18	365	0	0	0	0	0	383	95.30
Concrete road	7	30	412	0	0	0	0	449	91.94
Magma rock	0	0	0	335	3	0	18	356	94.10
Steel plate	0	0	0	0	335	8	34	377	88.86
Glass	0	0	0	0	0	201	10	211	95.26
Wall	0	0	0	3	5	5	398	411	96.84
Total	402	403	417	338	343	214	460
Accuracy (%)	93.78	90.57	98.80	99.11	88.86	93.93	86.52

Overall accuracy = 2423/2577 = 94.02%; kappa coefficient = (0.9402 − 0.1479)/(1 − 0.1479) = 0.93.