Author Affiliations
11. School of Mechanical and Electronic Engineering, Nanjing Forestry University, Nanjing 210037, China22. College of Food Science & Engineering of Nanjing University of Finance and Economics, Nanjing 210023, China33. Jiangsu Grain and Oil Quality Inspection Center, Nanjing 210031, Chinashow less
Fig. 1. Working principle of detection system
Fig. 2. Schematic diagram of device hardware
(a): Instrument strumcture diagram; (b): Physical drawing of instrument 1: Shell; 2: Vent; 3: Micro spectrometer; 4: Grip; 5: Upper cover and touch screen; 6: Raspberry Pi 4B; 7: Lug boss; 8: Light source control circuit board; 9: Card slot; 10: Testing accessories; 11: Power Supply
Fig. 3. Test accessories
1: Objective table; 2: Copper stud; 3: Light source seat; 4: Light source cover; 5: Sample pool; 6: Reflective cover; 7: Optical fiber collimating mirror; 8: Adjusting rod; 9: Swallowtail guide slide block; 10: Dovetail guide rail; 11: Support
Fig. 4. Spectra of flour samples with different quantities
Fig. 5. Light source control module
1: Circuit board; 2: Raspberry Pi 4B; 3: Lithium battery;4: 1 kΩ pull-up resistor; 5: Halogen lamp bead;6: TIP120 Darlington transistor
Fig. 6. Original spectra of flour
Fig. 7. Box plot graphics depicting the outliers in original spectra
Fig. 8. Box plot graphics depicting the outliers removed
Fig. 9. Savitzky-Golay smoothing
Fig. 10. Score plot of PCA
Fig. 11. PCA of modeling set samples
Fig. 12. Decision boundary
Fig. 13. Confusion matrix of LR prediction result
Fig. 14. LR ROC curves
Fig. 15. Testing system interface
Fig. 16. Work flow chart
Fig. 17. Test results of flour moisture
Yellow sample set | Min/% | Max/% | Average/% | Standard deviation/% |
---|
Calibrationset | 11.49 | 14.86 | 13.18 | 0.61 | Predictionset | 12.02 | 15.31 | 13.53 | 0.52 |
|
Table 1. Moisture measurement results of flour
Blue sample set | Min/(mg·kg-1) | Max/(mg·kg-1) | Average/(mg·kg-1) | Standard deviation /(mg·kg-1) |
---|
Calibration set | 0.005 1 | 8.455 6 | 0.873 4 | 1.339 6 | Prediction set | 0.021 2 | 8.999 6 | 0.856 1 | 1.518 6 |
|
Table 2. Measurement results of DON content in flour
Pre- treatment | Calibration set | Prediction set | RPD |
---|
| RMSEC/% | | RMSEC/% |
---|
None | 0.883 | 0.382 | 0.853 | 0.286 | 2.5 | MSC | 0.826 | 0.393 | 0.803 | 0.316 | 2.2 | SG | 0.872 | 0.351 | 0.794 | 0.264 | 2.4 | SNV | 0.783 | 0.410 | 0.722 | 0.323 | 2.0 |
|
Table 3. PLSR prediction models after different pretreatment
Number | Predict result | True result |
---|
01 | 1 | 1 | 02 | 0 | 0 | 03 | 0 | 0 | 04 | 0 | 0 | 05 | 1 | 1 | 06 | 0 | 0 | 07 | 1 | 1 | 08 | 0 | 0 | 09 | 1 | 0 | 10 | 0 | 0 | 11 | 0 | 0 | 12 | 1 | 1 | 13 | 1 | 1 | 14 | 0 | 0 | 15 | 0 | 0 | 16 | 0 | 0 | 17 | 1 | 0 | 18 | 0 | 0 | 19 | 0 | 0 | 20 | 0 | 0 |
|
Table 4. Detection result of samples in external verification set