Table 1. Performance of the PLSR model developed using hyperspectral data with various preprocessing at full wavelength and selected key wavelength

Wavelengthregion Pungencycomponent Pre-processing1) F2) Training model (n=315) Predictionmodel(n=135)
Calibration Cross-validation
Offsetc RMSEC Rc2 Offsetcv RMSECV Rcv2 Rp2
Full wavelength Capsaicin Raw 5 6.6863 33.0606 0.9776 6.8902 34.0150 0.9764 0.9700
Normalization 4 7.8969 35.9279 0.9735 8.1336 36.7714 0.9725 0.9631
SNV 3 9.4645 39.3325 0.9683 9.6388 40.0263 0.9674 0.9531
MSC 3 9.4645 39.3322 0.9683 9.6324 40.0252 0.9674 0.9531
SG 4 3.8072 24.9469 0.9872 4.2379 25.9852 0.9862 0.9796
Dihydrocapsaicin Raw 6 3.7256 12.0341 0.9769 4.0572 12.6012 0.9748 0.9719
Normalization 6 2.7254 10.2924 0.9831 2.9807 11.0459 0.9807 0.9807
SNV 6 1.3245 7.1757 0.9918 1.4575 7.6669 0.9907 0.9740
MSC 6 1.3231 7.1711 0.9918 1.4559 7.6572 0.9907 0.9736
SG 5 2.7787 10.3926 0.9828 3.2467 10.8770 0.9813 0.9685
Total capsaicinoid Raw 6 7.6009 37.0957 0.9835 8.0960 38.6849 0.9821 0.9833
Normalization 4 14.2431 50.7809 0.9690 14.7314 51.8528 0.9679 0.9568
SNV 4 11.1449 44.9196 0.9758 11.3963 46.2343 0.9745 0.9712
MSC 4 11.1453 44.9193 0.9758 11.3844 46.2284 0.9745 0.9713
SG 3 10.6840 43.9810 0.9768 12.1489 44.9624 0.9759 0.9649
Selected key wavelength Capsaicin Raw 5 17.7601 53.8793 0.9405 18.0993 54.8613 0.9387 0.9221
Normalization 3 21.0571 58.6663 0.9294 21.3966 59.5285 0.9278 0.9110
SNV 3 19.3931 56.3022 0.9350 19.7744 57.4230 0.9328 0.9241
MSC 3 19.3318 56.2113 0.9352 19.7084 57.3285 0.9331 0.9242
SG 3 17.6862 53.7674 0.9407 18.2175 54.7148 0.9390 0.9153
Dihydrocapsaicin Raw 4 10.0777 19.7921 0.9375 10.2481 20.1908 0.9354 0.9351
Normalization 3 10.4678 20.1710 0.9351 10.6284 20.4738 0.9336 0.9268
SNV 5 10.8211 20.5091 0.9329 11.3596 21.2718 0.9283 0.9123
MSC 5 11.0828 20.7554 0.9313 11.6233 21.5150 0.9266 0.9103
SG 3 9.3156 19.0289 0.9423 9.4943 19.3864 0.9404 0.9355
Total capsaicinoid Raw 4 28.1112 71.3358 0.9389 28.6070 72.6496 0.9370 0.9261
Normalization 3 32.2657 76.4280 0.9298 32.7568 77.5534 0.9282 0.9158
SNV 5 25.1292 67.4507 0.9453 26.2636 69.9567 0.9416 0.9354
MSC 3 35.1339 79.7534 0.9236 35.9736 81.4185 0.9209 0.9157
SG 3 26.4502 69.2007 0.9425 27.1553 70.3840 0.9409 0.9235
SNV, standard normal variate; MSC, multiplicative scatter correction; SG, Savitzky-Golay first derivative.
F, factor; RMSEC, root mean square errors of calibration; Rc2, correlation coefficient of the calibration; RMSECV, root mean square errors of validation; Rcv2, correlation coefficient of the full cross validation; Rp2, correlation coefficient of the prediction.