Back Article

Figure 1. 

Vis-NIR spectroscopy acquisition system.

Figure 2. 

Reference grape SSC tested on each day of the experiment.

Figure 3. 

Original spectral curves of grape SSC.

Figure 4. 

Grape SSC prediction result based on the whole band spectrum.

Figure 5. 

The grape SSC prediction result based on the SPA screening characteristic variables. (a) Performance comparison of predicting model for SSC from the SPA selected variables data with different pre-processing methods. (b) The distribution of variables selected by SPA.

Figure 6. 

The grape SSC prediction result based on the UVE screening characteristic variables. (a) Performance comparison of predicting model for SSC from the UVE selected variables data with different pre-processing methods. (b) The distribution of variables selected by UVE.

Figure 7. 

The grape SSC prediction result based on the CARS screening characteristic variables. (a) The RMSECV variation during the CARS algorithm iteration process. (b) The regression coefficient path of each variable with an increase in the number of sampling runs. (c) The performance comparison of predicting model for SSC from the CARS selected variables data with different pre-processing methods.(d) The distribution of variables selected by CARS.

Figure 8. 

Comparison of feature variables based on different selecting algorithms.

Figure 9. 

Correlation between SSC measurement values and predicted values in the calibration set and prediction set.