Objective To address issues such as the difficulty in accurate internal indicator quantification, the missed detection of external defects, and fragmented testing methods in the quality inspection of pickled products (taking duck eggs as an example), thereby enhancing the accuracy, efficiency, and intelligence of comprehensive food quality inspection in automatic production.Methods A comprehensive quality inspection method for pickled products is constructed based on an automatic food quality inspection system, which integrates deep learning, machine learning, and multiple detection technologies (hyperspectral technology + machine vision technology). Specifically, hyperspectral imaging technology is used to collect internal optical characteristic data of pickled duck eggs. After preprocessing, an improved partial least squares regression model is input to achieve synchronous detection of four core internal quality indicators: sodium chloride concentrations and moisture content in both egg yolk and egg white. Simultaneously, a machine vision system is employed to collect external image data of duck eggs. After preprocessing, the improved PP-YOLOE model is input to complete the recognition of surface crack defects and the rapid detection of yolk area ratio. The superiority of the proposed method is verified through comparative experiments.Results The experimental method achieves determination coefficients (R²) of greater than 0.975, and a root mean square error (RMSE) of less than 0.5% in internal indicator detection. The experimental method has an accuracy rate of 98.50% in crack defect identification, with a miss rate of only 0.40%. Additionally, the detection error of egg yolk area ratio is less than 1%, and the detection speed is increased by more than 20% compared to conventional methods.Conclusion This comprehensive quality inspection method for pickled products integrates multiple technologies, enabling integrated and high-precision detection of internal quality indicators and external defects/morphological parameters. It effectively overcomes the limitations of a single detection technology, consequently possessing strong engineering practicality and promotional value.