Apple Damage Detection Using Ethylene Gas Sensors

Abstract

To improve the automation and accuracy of postharvest apple quality grading and damage detection, and to overcome the inherent limitations of conventional manual inspection methods—most notably strong subjectivity, low operational efficiency, and the lack of support for online monitoring—this study develops an intelligent apple damage detection platform based on ethylene gas sensors. On this basis, a damage level identification approach is proposed that exploits the dynamic characteristics of ethylene concentration.

Apples of the same harvest season (Red Fuji) were selected as experimental samples and subjected to three treatments: undamaged, mildly damaged, and severely damaged. Ethylene concentration dynamics were continuously monitored over a five-day period under controlled temperature and humidity conditions. The results indicate that both the absolute levels and growth rates of ethylene concentration differ significantly across damage categories. Moreover, the average relative growth rate of ethylene concentration increases monotonically with the severity of damage.

Using the ethylene concentration growth rate as a key feature, a support vector machine (SVM) classification model was constructed to identify apple damage levels, and its performance was systematically compared with that of a recognition method based on image shape features. The experimental results demonstrate that, when ethylene growth rate features are employed, the SVM model achieves consistently high classification accuracy across multiple kernel functions. Among these, the radial basis function kernel exhibits the best performance, with an average recognition accuracy exceeding 97%, substantially outperforming approaches that rely solely on image-based features.

Overall, the findings provide strong evidence that apple internal damage can be identified with high precision by leveraging the dynamic behavior of ethylene concentration. This study thus offers a feasible and accurate solution for online postharvest apple quality monitoring and grading, and provides a valuable reference for nondestructive internal damage detection in agricultural products.