Discrimination of Fresh Tobacco Leaves with Different Maturity Levels by Near-Infrared (NIR) Spectroscopy and Deep Learning

Table 3

Discriminant accuracy (%) of different preprocessing methods.

Preprocessing methods	KNN	BPNN	SVM	ELM	CNN
Raw	55.33	77.05 ± 3.61	87.33	72.4 ± 3.25	92.35 ± 2.61
First derivation	85.33	88.32 ± 2.69	93.33	82.46 ± 4.44	95.84 ± 1.25
Second derivation	84.67	85.1 ± 2.54	92.67	80.24 ± 5.91	94.55 ± 1.65
SNV	74	86.67 ± 2.91	94	85.03 ± 3.43	94.36 ± 1.24
MSC	74	86.5 ± 2.52	93.33	84.49 ± 1.66	93.38 ± 1.42