EURASIP Journal on Advances in Signal Processing

Table 4 Classification accuracies for the MED classifications

From: Classification of hyperspectral imagery with neural networks: comparison to conventional tools

	194-band		30-band		13-band
	User	Producer	User	Producer	User	Producer
Class	Accuracy		Accuracy		Accuracy
A	98.39	98.39	98.39	98.71	98.39	98.71
B	96.05	36.50	56.76	21.00	66.67	23.00
C	54.63	56.50	42.80	55.00	41.37	51.50
D	91.44	96.67	92.69	96.67	90.67	97.14
E	88.20	67.62	86.55	70.48	87.10	64.29
F	41.05	97.50	40.45	90.00	41.30	95.00
G	100.00	67.05	100.00	65.32	100.00	65.32
H	98.14	95.91	97.42	85.91	98.48	88.64
I	33.49	70.00	24.62	65.00	24.33	64.00
J	100.00	100.00	98.02	98.80	98.81	99.60
K	97.22	84.00	99.06	84.00	97.98	77.60
L	93.69	97.20	93.24	96.73	93.27	97.20
M	98.18	27.00	91.61	16.50	88.24	22.50
N	97.66	83.33	30.11	9.33	61.88	33.00
O	88.86	95.16	60.10	78.71	71.79	81.29
P	80.00	96.00	66.14	97.67	65.31	96.00
Q	99.07	96.82	97.69	95.91	96.76	95.00
R	100.00	94.00	100.00	92.00	93.33	84.00
S	81.94	98.33	67.06	95.00	65.48	91.67
T	45.61	80.00	54.64	81.54	48.21	83.08
U	43.21	78.00	43.73	68.00	33.89	51.00
V	71.43	70.00	47.83	66.00	43.06	62.00
W	95.40	83.00	96.05	73.00	94.70	71.50
	Accuracy (%) = 82.04		Accuracy (%) = 72.85		Accuracy (%) = 73.29
	κ = 0.8107		κ = 0.7140		κ = 0.7187

Back to article page