A Search Method for Optimal Band Combination of Hyperspectral Imagery Based on Two Layers Selection Strategy

Table 3

Classification accuracies (%) achieved using 20 selected bands in PaviaU with SVM classifier.

Class	Algorithm
	Accuracy (%)
	MVPCA	WaLuDi	DBSCAN	FDPC	LP	ISSC	TLS
Asphalt	86.5	87.6	87.7	89.8	88.6	89.7	91.4
Meadows	96.4	96.8	96.2	94.8	94.7	98.1	96.6
Gravel	72.7	71.3	75.8	74.2	73.7	74.6	76.2
Trees	84.2	91.3	91	92.1	90.3	92.2	93.3
Painted metal sheets	99.1	99.4	99.4	97.8	97.3	98.2	99.1
Bare soil	53.2	82.5	81.5	82.4	80.4	82.8	83.5
Bitumen	79.8	83.8	81.7	76.6	80.1	82.7	83.3
Self-blocking bricks	83.4	85.2	86.3	83.8	82.4	83.6	84.9
Shadows	97.2	98.5	98.5	98	97.8	98.3	98.7
AA	83.6	88.4	88.6	87.7	87.3	88.9	89.6
OA	85.4	89.3	89.1	89.3	87.6	89.7	91.5
Variance	183.9	73.5	60.7	70.1	64.9	64.4	57.5