TABLE 3.

The classification results based on PCA/AE features of different spectral sources.

Cultivar	Features	Variables	Model	Tr (%)	Val (%)	Te (%)	Model	Tr (%)	Val (%)	Te (%)	Model	Tr (%)	Val (%)	Te (%)
Zhefujing83	PCA-HSI	14	SVM	98.33	95.00	100	LR	98.33	100	100	CNN	100	95.00	94.23
	PCA-MIR	34		100	100	90.38		100	100	90.38		100	95.00	90.38
	PCA-LIBS	114		100	85.00	63.46		100	85.00	63.46		100	95.00	78.85
	AE-HSI	32		98.33	70.00	76.92		98.33	70.00	65.38		98.33	80.00	63.4
	AE-MIR	16		94.17	80.00	84.38		96.67	95.00	76.92		100	100	78.85
	AE-LIBS	64		99.17	95.00	63.46		99.17	90.00	65.38		100	100	69.23
AD516	PCA-HSI	14		100	100	96.88		100	100	100		100	100	93.75
	PCA-MIR	42		100	100	96.88		100	100	95.31		100	95.00	78.13
	PCA-LIBS	103		100	90.00	81.25		100	95.00	81.25		95.83	95.00	85.94
	AE-HSI	32		94.17	75.00	81.25		97.5	95.00	84.38		95.00	85.00	84.38
	AE-MIR	16		96.67	80.00	87.50		95.00	85.00	84.38		99.17	85.00	78.13
	AE-LIBS	64		100	90.00	73.44		100	95.00	81.25		100	95.00	73.44

Other abbreviations are similar to this fashion. Tr, training set; Val, validation set; Te, test set; PCA-HSI, the features of HSI extracted by PCA; AE-HSI, the features of HSI extracted by AE.