3.3.2. Determination of characteristic parameters for detecting
scattered yolk
Based on morphological characteristic parameters, scattered yolk was classified using SVC model (Table 4). Accuracy was
98.3% in the calibration set, and 96.3% in the prediction set,respectively. Wang et al. (2012) used computer vision to detect double yolk eggs from single yolk eggs, and the yolk morphology was used. In this study, the egg yolk membrane rupture would lead to the diffuse of yolk morphology, and hyperspectral images could reflect this change and thus the yolk morphology provided useful information for internal quality detection of eggs. In another way, although identification of scattered yolk using this non-destructive approach is promising for the application in the egg industry, these results would be better interpreted with caution since scattered yolk was generated in our laboratory by the simulation of transportation vibration. The accuracy and robustness of models could be further improved with larger sample size and actual situations.