Spectral-Spatial Hyperspectral Image Classification Based on Homogeneous Minimum Spanning Forest

Table 4

Information classes, number of training samples, and classification accuracy for PaviaU.

	No. of samples	No. of training samples	SVM	SVM-MSF	MC-MSF	RD-MSF	MSWMF-MSF	MSMMSF	Proposed Methods
									HMSF	HKNN-MSF-MV	F–HKNN-MSF-MV
Asphalt	6631	548	82.17	93.16	87.1	93.85	97.42	99.2	93045	93.16	97.3
Meadows	18649	540	87.25	97.7	96.67	98.1	98.53	99.23	99.38	99.45	100
Gravel	2099	392	81.23	92.15	78.37	93.79	91.33	95.43	97.28	98.2	98.5
Trees	3064	524	93.21	91.24	98.87	94.38	94.68	96.28	77.58	88.84	98.2
Metal sheets	1345	265	99.78	96.3	96.8	96.9	98.36	100	99.78	99.78	100
Bare soil	5029	532	91.05	99.91	96.4	98.5	98.35	99.8	99.52	99.91	99.2
Bitumen	1330	375	89.4	98.57	98.87	98.9	99.25	100	99.25	97.8	98.87
Bricks	3682	514	85.17	92.78	90.10	91.5	96.28	98.18	82.62	87.6	96.86
Shadows	947	231	99.89	96.23	98.62	97.36	96.6	97.37	99.89	99.89	100
AA (%)	—	—	89.9	95.33	93.52	95.92	96.75	98.36	94.37	96.07	98.77
OA (%)			87.6	96.67	95	97.07	97.45	98.8	95.65	97.23	99.12
Kappa coefficient (%)			85.33	94.19	92.83	96.05	96.64	98.41	94.21	96.31	98.92