A gas information adaptive deep learning network combined with an electronic nose to identify the egg quality at different storage periods

Abstract

Background

As the storage period of eggs extends, their quality declines significantly, highlighting the need for a fast and efficient method for assessing egg quality. This study introduces a Deep Gas Features Calculation Network (DGFCNet) integrated with an electronic nose (e-nose) system to enable accurate egg quality identification across various storage periods.

Methods

The e-nose system, equipped with an array of gas sensors, gathers gas profile data from eggs at different storage periods. A Deep Gas Features Calculation Module (DGFCM) is developed to extract deep gas features that enhance classification result. This module reduces computational complexity through grouped calculations, enhances the representation of deep features via an attention mechanism, and adaptively integrates shallow and deep features with residual dense connections to prevent feature degradation. Based on DGFCM, DGFCNet is designed to identify the gas information of egg.

Significant Findings

DGFCNet achieves egg quality identification at different storage times, reaching 97.30 % accuracy, 97.56 % precision, and 97.47 % recall, surpassing other leading methods in gas information classification. In conclusion, DGFCNet combined with the e-nose system offers an effective technological approach for monitoring egg quality throughout storage.