Table 5.

Classification results for SVM, k-NN, and RBFNN models based on optimal wavelengths selected by PCA.

	Model	Parameter 5	Calibration Set			Prediction Set
			Acc. 6 (%)	Sen. 7	Spe. 8	Acc. (%)	Sen.	Spe.
Pixel to pixel 1	SVM	(256, 5.28)	91.83	0.898	0.939	80.10	0.800	0.802
	k-NN	3	78.48	0.700	0.870	78.18	0.642	0.895
	RBFNN	7	88.40	0.842	0.926	80.89	0.797	0.819
Pixel to object 2	SVM	(256, 5.28)	91.83	0.898	0.939	93.62	0.785	0.998
	k-NN	3	78.48	0.700	0.870	83.82	0.464	0.992
	RBFNN	7	88.40	0.842	0.926	91.40	0.711	0.997
Object to pixel 3	SVM	(147, 9.12)	99.72	0.994	0.998	71.10	0.817	0.626
	k-NN	5	95.46	0.870	0.991	76.86	0.727	0.803
	RBFNN	3	99.78	0.994	0.999	54.14	0.819	0.317
Object to object 4	SVM	(147, 9.12)	99.72	0.994	0.998	99.12	0.987	0.993
	k-NN	5	95.46	0.870	0.991	94.06	0.839	0.982
	RBFNN	3	99.78	0.994	0.999	99.30	0.983	0.997

¹ Pixel to pixel means to use models using pixel-wise spectra to predict pixel-wise spectra; ² Pixel to object means models using pixel-wise spectra to predict object-wise spectra; ³ Object to pixel means to use models using object-wise spectra to predict pixel-wise spectra; ⁴ Object to object means to use models using object-wise spectra to predict object-wise spectra; ⁵ Parameters for SVM models are C and γ, parameter for k-NN is number of neighbors (k) and parameter for RBFNN is spread value; ⁶ Acc. means accuracy; ⁷ Sen. means sensitivity; ⁸ Spe. means specificity.