Prediction and visualisation of S-ovalbumin content in egg whites using hyperspectral images

Figures & data

Figure 1. Schematic diagram of transmission spectrum image acquisition device, 1: Computer, 2: USB line, 3: High spectrometer, 4: CCD camera, 5: Compact black box, 6: Egg, 7: Light box, 8: Mobile platform, and 9: Light

Table 1. Sample set of S-ovalbumin content

Samples	Maximum	Minimum	Mean	Standard deviation
Total sample (108)	106.2%	12.8%	63.1%	0.279
Calibration set (72)	106.2%	21.1%	65.1%	0.223
Validation set (36)	105.1%	12.8%	52.6%	0.274

Figure 2. The regularity of S-ovalbumin content change with day

Table 2. Modelling effect of S-ovalbumin protein content in different calibration models

Pretreatment method	PLSR
	RC	RMSEC	RP	RMSEP	RPD
None	0.868	0.110	0.841	0.146	1.578
FD	0.849	0.117	0.848	0.143	1.626
SD	0.822	0.126	0.832	0.150	1.525
SG	0.868	0.110	0.841	0.146	1.578
SNV	0.929	0.082	0.875	0.150	2.082
MSC	0.929	0.082	0.874	0.242	1.592
Normalisation	0.899	0.097	0.910	0.112	2.224

Figure 3. SPA extraction of characteristic wavelengths: (a) changes in RMSE in the model and (b) optimal characteristic wavelength selected by SPA

Table 3. The optimal wavelength combinations based on the SPA method

Index	Wavelength (nm)
S-ovalbumin content	460.29; 622.34; 670.07; 709.01; 732.88; 759.26; 855.98; 902.46; 916.27; 961.49; 964.01; 997.92

Table 4. Model effect of S-oval protein content based on characteristic wavelength

	PLSR					MLR
Modelling variables (number)	RC	RMSEC	RP	RMSEP	RPD	RC	RMSEC	RP	RMSEP	RPD
Normalised +SPA (12)	0.880	0.105	0.867	0.143	1.866	0.922	0.086	0.911	0.119	2.348

Figure 4. Correlation between measured and predicted values of S-ovalbumin content

Figure 5. Visualisation of S-ovalbumin content distribution in three egg samples