Electronic Nose and Deep Learning Approach in Identifying Ripe Lycopersicum esculentum L. TomatoFruit

Identification techniques that determine the ripeness of fruits already have several approaches, but most of the methods use instrumental methods that require sophisticated equipment operated by trained technicians. Electronic nose systems have already been used to identify the freshness and ripeness of other fruits, including tomatoes. This is due to the own volatile organic compound profiles of fruits in various stages of maturity. In this study, an electronic nose system is made using a gas sensor array, Raspberry Pi 3B, and a deep learning model. The model is trained using 150 samples wherein data augmentation is implemented to produce 750 grayscale image patterns. The tomatoes are classified into three: ripe, not ripe, and unknown. During training, the model that undergoes data augmentation and uses 24×8 input data has the highest mean score, 0.96. This model is used in the electronic nose system, and 43 out of 50 samples are correctly identified, which is an 86% accuracy.