Scholarlink

<p>Many kinds of volatile organic compounds (VOCs) excrete from fruits when they are stored, which can be used for freshness evaluation. Due to this characteristic, in this study, VOCs samples are sampled from headspace of apple pieces with our self-developed electronic nose (e-nose) system. A simple feature selection method, called three phase feature integration (TPFI) was designed to obtain features of e-nose signals in this paper. With our method, 4 traditional feature engineering methods were combined and tested with 5 machine learning models. Our results showed that, TPFI can optimize single feature extraction method, even in the same dimension. In the study of recognizing 6 stages of apple spoilage, TPFI - PCA outperforms other united feature extraction methods, which achieving 99.6% accuracy and 0.993 F1 score under Multilayer Perceptron (MLP) model. Our findings suggested fruits spoilage can be accurately evaluated with our e-nose system in different storage periods, and TPFI feature extraction methods could improve its performance further when using traditional feature extraction methods. As a reliable and contactless technology, e-nose system showed significant potential on monitoring the shelf-life for real-world applications in the food industry.</p>

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

Evaluate of apple spoilage through TPFI using self-developed electronic nose system

  • 期刊论文

  • 发表日期：

    2025年12月12日

查看全文
  • Yimin Zhou,
  • Jie Min,
  • Hengyu Zhou,
  • Pengxin Li,
  • Jiajie Xu,
  • Yirong Wang,
  • Wei Liang,
  • Wang Li,
  • Gen Li

摘要

Many kinds of volatile organic compounds (VOCs) excrete from fruits when they are stored, which can be used for freshness evaluation. Due to this characteristic, in this study, VOCs samples are sampled from headspace of apple pieces with our self-developed electronic nose (e-nose) system. A simple feature selection method, called three phase feature integration (TPFI) was designed to obtain features of e-nose signals in this paper. With our method, 4 traditional feature engineering methods were combined and tested with 5 machine learning models. Our results showed that, TPFI can optimize single feature extraction method, even in the same dimension. In the study of recognizing 6 stages of apple spoilage, TPFI - PCA outperforms other united feature extraction methods, which achieving 99.6% accuracy and 0.993 F1 score under Multilayer Perceptron (MLP) model. Our findings suggested fruits spoilage can be accurately evaluated with our e-nose system in different storage periods, and TPFI feature extraction methods could improve its performance further when using traditional feature extraction methods. As a reliable and contactless technology, e-nose system showed significant potential on monitoring the shelf-life for real-world applications in the food industry.