Ultrasound/electrostatic field combined with ε-polylysine for improving the quality of ready-to-eat pork jelly and freshness classification based on convolutional neural network.

This study investigated the effects of ultrasound (US, 300 W, 20 kHz, 20 min) and electrostatic field (EF, 3.5 kV/cm, 60 min) combined with ε-polylysine (PL, 0.5%, w/w) on the quality of ready-to-eat pork jelly (Shuijingyaorou, SJYR) during 40-day storage at 4 °C. Physicochemical properties such as lipid oxidation, total plate count, texture, flavor, and moisture distribution were evaluated. The results showed that the combined treatment (US + EF + ε-polylysine) exhibited the strongest preservation effect for SJYR, significantly inhibiting lipid oxidation, protein degradation, pH increase, and microbial growth, meanwhile maintaining texture and flavor of SJYR, resulting in the shelf-life extended approximately 10 days compared with the blank control group. Furthermore, four convolutional neural network (CNN) models including Xception, ResNet-50, MobileNet, and GhostNet were developed for SJYR freshness classification. Xception model achieved the highest test accuracy (83.05% mean precision), while GhostNet model showed promising accuracy (77.48%) with lower computational cost, making it suitable for mobile applications. This study provides a feasible strategy for improving SJYR shelf-life and rapid freshness identification.