Scholarlink

Electronic noses have been the subject of several research studies in the agricultural sector to develop quality control tools that manage the ripeness of fruits. Its importance lies in the capabilities of these devices to perform classification tasks based on the emission of volatile organic compounds (VOCs), which is especially useful when dealing with climacteric fruits. In the present study, an electronic nose was developed to classify the ripeness of the Musa Cavendish banana according to the Von Loesecke scale by training a Multilayer Perceptron (MLP) algorithm with the derivative of the signals acquired by the sensor array. The results show that only the first 50 s of signal acquisition are needed to achieve an accuracy of 93.42%. These results surpass those of algorithms such as K-Nearest Neighbours (KNN) and Support Vector Machine (SVM). The importance of each sensor in the classifier model was also investigated by performing a Spearman correlation test so that the amount required by the electronic sense of smell to operate correctly could be reduced.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Electronic Nose for Classification of Banana Ripeness by Using MLP, KNN, and SVM Algorithms

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Victor A. Vargas Alarcón,
  • Samir E. Parra Ricciulli,
  • Kelvin Beleño Saenz,
  • Carlos G. Diaz Saenz,
  • José S. Torrecilla Velasco,
  • Javier Jiménez-Cabas,
  • Luis Diaz-Charris

摘要

Electronic noses have been the subject of several research studies in the agricultural sector to develop quality control tools that manage the ripeness of fruits. Its importance lies in the capabilities of these devices to perform classification tasks based on the emission of volatile organic compounds (VOCs), which is especially useful when dealing with climacteric fruits. In the present study, an electronic nose was developed to classify the ripeness of the Musa Cavendish banana according to the Von Loesecke scale by training a Multilayer Perceptron (MLP) algorithm with the derivative of the signals acquired by the sensor array. The results show that only the first 50 s of signal acquisition are needed to achieve an accuracy of 93.42%. These results surpass those of algorithms such as K-Nearest Neighbours (KNN) and Support Vector Machine (SVM). The importance of each sensor in the classifier model was also investigated by performing a Spearman correlation test so that the amount required by the electronic sense of smell to operate correctly could be reduced.