Quantitative Determination of Germinability of Puccinia striiformis f. sp. tritici Urediospores Using Near Infrared Spectroscopy Technology

Table 4

The prediction results of the optimal SVR models using different preprocessing methods of the near infrared spectra for quantitative determination of germinability of Puccinia striiformis f. sp. tritici urediospores.

Preprocessing method	The ratio of training set to testing set	Spectral region/cm^−1	Optimal parameters		Training set		Testing set
						MSE		MSE
No preprocessing	3 : 1	4000–12000	3.0314	5.278	0.9475	0.002277	0.8996	0.005421
Level 1 decomposition denoising using db2 wavelet	3 : 1	4000–12000	3.0314	5.278	0.9464	0.002322	0.8996	0.005422
Level 2 decomposition denoising using db2 wavelet	3 : 1	4000–12000	3.0314	5.278	0.9429	0.002474	0.8978	0.005508
Level 3 decomposition denoising using db2 wavelet	3 : 1	4000–12000	3.0314	5.278	0.9393	0.002630	0.8954	0.005574
Euclidean normalization	3 : 1	7000–12000	84.4485	147.0334	0.9047	0.004231	0.8956	0.005408
Multiplication scatter correction	3 : 1	7000–12000	48.5029	1	0.8925	0.004805	0.8754	0.007024
Standard normalized variate transform	3 : 1	6000–9000	84.4485	0.0359	0.8531	0.006267	0.9065	0.007758
Vector normalization	2 : 1	4000–12000	147.0334	0.3299	0.9340	0.003024	0.9034	0.004604
	5 : 1	8000–11000	27.8576	16	0.9461	0.002464	0.9262	0.003488
	3 : 1	8000–12000	256	0.3299	0.9320	0.003128	0.9129	0.007432
	5 : 1	8000–12000	147.0334	0.5743	0.9462	0.002639	0.9259	0.004726
S-G first derivative transform	3 : 1	5000–9000	147.0334	256	0.9324	0.002947	0.8817	0.007193
S-G second derivative transform	5 : 1	8000–10000	256	256	0.9099	0.004248	0.8519	0.010270