Optimal Band Configuration for the Roof Surface Characterization Using Hyperspectral and LiDAR Imaging

Table 1

Classification accuracies.

Compression method	Group	Bands used	Classifier	OA (%)	Kappa	RMSE-ANN	PA (%)	UA (%)
Number	303	303	SAM	60.17	0.57		65.83	64.08
			ANN	83.73	0.82	0.42	74.94	79.64
			SVM	89.72	0.88		89.21	90.27
BS-PCALo	76	76	SAM	57.92	0.55		63.54	62.72
			ANN	94.19	0.93	0.21	95.21	95.03
			SVM	87.90	0.86		88.00	89.17
		48	SAM	56.12	0.53		61.36	62.05
			ANN	93.18	0.92	0.25	93.65	94.24
			SVM	87.99	0.88		87.57	88.77
	48	37	SAM	56.13	0.53		61.44	62.36
			ANN	92.26	0.91	0.27	93.19	92.76
			SVM	87.93	0.87		87.53	88.58
		20	SAM	56.85	0.54		62.05	63.26
			ANN	91.85	0.9	0.29	92.05	92.50
			SVM	87.9	0.86		87.64	88.63
		10	SAM	57.7	0.54		62.66	62.82
			ANN	90.85	0.90	0.4	91.86	91.41
			SVM	86.87	0.85		86.36	87.60
		6	SAM	51.56	0.48		58.95	56.46
			ANN	77.12	0.70	0.62	73.61	78.43
			SVM	81.99	0.80		81.63	81.67
		4	SAM	44.35	0.40		51.73	46.67
			ANN	59.77	0.56	0.75	57.58	58.33
			SVM	69.23	0.66		66.23	72.40
		15 (5 V + 5 NIR + 5 MIR)	SAM	56.22	0.53		61.87	61.95
			ANN	90.48	0.89	0.34	91.00	91.20
			SVM	86.99	0.85		86.62	87.78
		10 (5 V + 5 NIR)	SAM	57.31	0.54		63.04	60.91
			ANN	87.07	0.86	0.39	84.32	86.81
			SVM	86.97	0.85		86.20	87.55
		10 (5 N + 5 MIR)	SAM	41.32	0.37		47.36	39.82
			ANN	77.79	0.76	0.59	75.34	76.56
			SVM	70.1	0.67		66.54	68.07
		10 (5 V + 5 MIR)	SAM	52.47	0.49		58.35	58.71
			ANN	89.97	0.89	0.38	89.48	91.76
			SVM	86.09	0.85		85.78	87.21
MNF transform	303	14 MNF bands	SAM	Failed	Failed		0.00	0.00
			ANN	79.78	0.78	0.41	69.63	72.68
			SVM	86.13	0.84		84.89	85.98
PCA transform		14 PCA bands	SAM	59.12	0.56		64.84	65.35
			ANN	35.25	0.29	0.9	36.55	29.78
			SVM	91.94	0.91		92.10	92.40