Author Affiliations
11. College of Mechanical and Electrical Engineering, Sichuan Agricultural University, Ya'an 625014, China22. College of Information Engineering, Sichuan Agricultural University, Ya'an 625014, Chinashow less
Fig. 1. Hyperspectral sorter
Fig. 2. Spectral information of kiwi samples
Fig. 3. Spectral reflectance images before and after DOSC preprocessing
Fig. 4. The correlation coefficient between spectral band and sugar content before and after DOSC preprocessing
Fig. 5. Extracting characteristic variables by IRIV
Fig. 6. Distribution of IRIV characteristic spectral variables
Fig. 7. The results of CARS
Fig. 8. Distribution diagram of SPA spectral characteristic variables
Fig. 9. Distribution of SPA characteristic spectral variables for kiwifruit sugar content
Data set | Samples | Min | Max | Mean | S.D |
---|
Calibration set | 90 | 9.50 | 18.10 | 13.17 | 1.93 | Prediction set | 30 | 10.40 | 16.50 | 13.56 | 1.52 |
|
Table 1. The statistical results of kiwi fruit sugar content measurement (unit: /°Brix)
Pretreatment method | Number of hidden neurons | Calibration set | Prediction set |
---|
RMSEC | RC | RMSEP | RP |
---|
Raw valid data | 92 | 1.076 2 | 0.673 4 | 1.126 0 | 0.539 9 | SNV | 72 | 1.418 2 | 0.410 2 | 1.414 9 | 0.342 1 | MSC | 71 | 1.345 7 | 0.469 0 | 1.154 7 | 0.561 8 | DOSC | 35 | 0.518 4 | 0.927 3 | 0.753 2 | 0.744 9 |
|
Table 2. All-band ELM prediction model using different pretreatment methods
Extraction method | Variable number |
---|
CARS | 49 | SPA | 9 | IRIV | 8 | CARS+SPA | 58 | CARS+IRIV | 55 | (CARS+SPA)-SPA | 11 | (CARS+IRIV)-SPA | 19 |
|
Table 3. The number of feature variables extracted by different feature extraction methods
Extraction method | Variable number | γ | σ2 |
---|
CARS | 49 | 178 232.022 6 | 2 315.787 8 | SPA | 9 | 24 689.722 8 | 4 419.308 9 | IRIV | 8 | 68 345.149 3 | 4 095.753 5 | CARS+SPA | 58 | 128 353.819 6 | 519.357 2 | CARS+IRIV | 55 | 66 321.544 4 | 3 244.478 3 | (CARS+SPA)-SPA | 11 | 213 051.988 1 | 2 541.380 7 | (CARS+IRIV)-SPA | 19 | 292 698.908 2 | 810.048 4 |
|
Table 4. PSO optimized parameters γ, σ2
Extraction method | Variable number | RMSEC | RC | RMSEP | RP | RPD |
---|
CARS | 49 | 0.658 3 | 0.882 7 | 0.878 1 | 0.653 3 | 1.727 4 | SPA | 9 | 0.672 6 | 0.877 6 | 0.881 6 | 0.650 6 | 1.720 6 | IRIV | 8 | 0.652 5 | 0.884 8 | 0.858 2 | 0.668 8 | 1.767 4 | CARS+SPA | 58 | 0.659 5 | 0.882 3 | 0.886 6 | 0.646 6 | 1.710 9 | CARS+IRIV | 55 | 0.654 2 | 0.884 2 | 0.871 6 | 0.658 4 | 1.740 3 | (CARS+SPA)-SPA | 11 | 0.662 5 | 0.881 2 | 0.895 9 | 0.639 1 | 1.693 1 | (CARS+IRIV)-SPA | 19 | 0.666 6 | 0.879 7 | 0.883 1 | 0.649 4 | 1.717 7 |
|
Table 5. Prediction results of SVR model established by different feature extraction methods
Extraction method | Variable number | RMSEC | RC | RMSEP | RP | RPD |
---|
CARS | 49 | 0.451 5 | 0.944 8 | 0.834 8 | 0.686 7 | 1.817 0 | SPA | 9 | 0.637 3 | 0.890 1 | 0.874 5 | 0.656 2 | 1.734 5 | IRIV | 8 | 0.624 0 | 0.894 6 | 0.867 1 | 0.662 0 | 1.749 3 | CARS+SPA | 58 | 0.138 7 | 0.994 8 | 0.823 1 | 0.695 4 | 1.842 8 | CARS+IRIV | 55 | 0.471 9 | 0.939 7 | 0.824 8 | 0.694 1 | 1.839 0 | (CARS+SPA)-SPA | 11 | 0.580 5 | 0.908 8 | 0.870 7 | 0.659 1 | 1.742 1 | (CARS+IRIV)-SPA | 19 | 0.471 5 | 0.939 8 | 0.788 5 | 0.720 4 | 1.923 5 |
|
Table 6. Prediction results of LSSVM model established by different feature extraction methods
Extraction method | Variable number | Number of hidden neurons | RMSEC | RC | RMSEP | RP | RPD |
---|
CARS | 49 | 46 | 0.449 7 | 0.945 3 | 0.685 0 | 0.789 0 | 2.214 4 | SPA | 9 | 89 | 0.585 4 | 0.907 2 | 0.799 3 | 0.712 7 | 1.897 7 | IRIV | 8 | 86 | 0.557 3 | 0.916 0 | 0.796 8 | 0.714 5 | 1.903 6 | CARS+SPA | 58 | 50 | 0.466 6 | 0.941 1 | 0.637 2 | 0.817 4 | 2.380 4 | CARS+IRIV | 55 | 50 | 0.449 2 | 0.945 4 | 0.591 8 | 0.842 5 | 2.563 0 | (CARS+SPA)-SPA | 11 | 49 | 0.511 8 | 0.929 1 | 0.688 6 | 0.786 8 | 2.202 6 | (CARS+IRIV)-SPA | 19 | 92 | 0.450 3 | 0.945 1 | 0.598 3 | 0.839 0 | 2.535 1 |
|
Table 7. Prediction results of ELM model established by different feature extraction methods