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

Table 4

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

Class	Algorithm
	Accuracy (%)
	MVPCA	WaLuDi	DBSCAN	FDPC	LP	ISSC	TLS
Brocoli-green-weeds-1	83.4	84.5	83.2	82.7	82	84.4	84.7
Brocoli-green-weeds-2	88.8	92.3	90.2	94.2	89	91.7	91.4
Fallow	80.5	81.4	83.3	83.6	81.4	84.2	84.8
Fallow-rough-plow	76.4	80.5	79.2	77.8	79.6	80.4	81.1
Fallow-smooth	86.7	86.2	91.6	85	87.3	86.8	86.4
Stubble	92.8	93.1	89.6	91.6	89.2	90.4	88.7
Celery	87.7	89.1	92.4	92.8	90.3	92.2	87.4
Grapes-untrained	95.4	96.1	97.3	97	95.7	95.4	96.5
Soil-vinyard-develop	92.6	93.5	92.2	81.8	86.2	93.4	91.7
Corn-seneseed-green-weeds	86.9	89.5	86.2	90.3	84.2	86.3	87.7
Lettuce-romaine-4wk	75.4	79.2	78.5	80.4	75	79.3	79.7
Lettuce-romaine-5wk	81.6	83.2	83	84.4	81.2	81.6	84.8
Lettuce-romaine-6wk	74	72.6	72.3	76.2	73.6	75.3	75.5
Lettuce-romaine-7wk	73.2	76.3	74.8	75.5	74.6	75	77.4
Vinyard-untrained	95	95.3	94.7	95.2	91.2	95.6	96.2
Vinyard-vertical-trellis	80.4	80.7	81.7	81.6	76.7	81.5	82.3
AA	84.4	85.8	85.6	85.6	83.6	85.8	86.0
OA	87.5	89.2	89.1	89.8	88	89.2	90.8
Variance	52.0	47.9	50.1	45.6	41.4	42.7	34.7