Classification results of RBFNN models developed on optimal wavelengths with 3000 and 1100 samples of each variety in the calibration set.

Samples of each variety in the calibration set		Calibration				Prediction
		1	2	3	Accuracy	1	2	3	Accuracy
3000	1	2799	48	153	93.30%	1176	49	75	90.46%
	2	44	2907	49	96.90%	40	1196	64	92.00%
	3	166	65	2769	92.30%	83	37	1180	90.77%
	Total				94.17%				91.08%
1100	1	1010	20	70	91.82%	1184	49	67	91.08%
	2	19	1063	18	96.64%	46	1195	59	91.92%
	3	66	10	1024	93.09%	87	43	1170	90.00%
	Total				93.85%				91.00%