Аңдатпа
Reducing post-harvest losses requires timely, non-destructive methods for assessing the freshness of stored crops. While electronic noses (e-noses) and machine learning have been explored for fruits and high-value produce, there is limited research on staple root vegetables and the impact of different storage conditions. This study presents a multisensory e-nose combined with a feedforward neural network to classify the freshness of potatoes and carrots stored under humid and dry environments for six months. Gas-emission profiles of key volatiles such as carbon dioxide and ammonia were recorded and used to label samples as fresh, mildly spoiled, or spoiled. The classifier achieved accuracies of up to 90% for humid-stored potatoes but lower performance (~70%) for dry-stored carrots, where rapid spoilage produced overlapping volatile signatures. These results demonstrate the feasibility of integrating low-cost e-noses with machine learning for freshness monitoring, while also highlighting limitations such as small sample size, single laboratory setting, and challenges in transitional spoilage states. Future work will focus on expanding datasets, testing additional volatile sensors, and improving generalizability for deployment in real storage environments.