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

Table 2

Classification accuracies (%) achieved using 40 selected bands in Indian Pines with KNN classifier.

Class	Algorithm
	Accuracy (%)
	MVPCA	WaLuDi	DBSCAN	FDPC	LP	ISSC	TLS
Alfalfa	66.8	54.6	50.5	39.2	33.7	34.2	45.4
Corn-notill	60.5	71.3	69.6	71	64.3	75.8	73.7
Corn-mintill	41.5	57.4	60	66.8	48.7	65.2	66.8
Corn	48.4	57.4	57.7	58.3	53.8	53.1	60.1
Grass-pasture	82.6	88.7	86.2	85	88.4	82.1	84.4
Grass-trees	93.1	94.5	95	94.3	93.3	96.7	96
Grass-pasture-mowed	45.2	44.8	42.1	37.3	71.7	92	45.2
Hay-windrowed	97.6	96.3	96.6	95.8	98.1	96.7	96.6
Oats	29.8	24.2	27	24.4	38.3	61.2	29.7
Soybean-notill	59.5	76.2	73.7	68.3	67.2	79.4	75.7
Soybean-mintill	77.4	79.3	81.5	82.6	82.9	82.5	84.2
Soybean-clean	43.1	69.6	71	72.2	58.9	83.3	72
Wheat	93	97.7	98.4	98.1	96.2	96.2	95.4
Woods	93.6	94.4	93.3	93.5	94.6	96.8	95.8
Buildings-grass-trees-drives	42.7	52.7	51.9	50.4	54.9	53.6	55.1
Stone-steel-towers	94.9	76.3	78.4	83.6	69.7	80.9	82.4
AA	66.8	70.9	70.6	70	69.7	76.9	72.4
OA	74	77.8	76.2	77.1	75.3	79.7	80.8
Variance	515.4	421.4	421.1	487.7	410.0	326.2	399.3