Scholarlink

<p>The geographical origin of rice significantly impacts its nutritional composition, flavor, and price. Traditional rice traceability detection methods face challenges such as complex detection procedures, high equipment costs, and time inefficiency. This study proposes a rapid detection method for rice origin based on an electronic tongue (ET) combined with a meta-learning and contrastive learning dual aggregation network (MCDANet). Firstly, the taste response signals of rice samples from different origins are collected by an ET device. A dual aggregation model is proposed to analyze the collected ET signals. This model employs a prototypical network based on meta-learning as the inner model for feature extraction and pattern recognition of ET signals. To enhance the feature extraction capability of the inner model, a SimSiam network based on contrastive learning is incorporated as the outer model to optimize the prototypical network. Finally, the proposed model is trained with a semi-supervised strategy under few sample conditions. The experimental results demonstrate that this method achieves superior recognition accuracy in classifying ET signals of different rice samples, with an overall accuracy of 98%. This study provides a novel and rapid detection method for rice origin traceability, which has a promising application in agriculture and 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

Application of electronic tongue combined with meta-learning and contrastive learning dual aggregation network for rapid detection of rice origins

  • 期刊论文

  • 发表日期：

    2025年12月27日

查看全文
  • Xin Li,
  • Zhiqiang Wang,
  • Jingbao Wang,
  • Yubin Lan

摘要

The geographical origin of rice significantly impacts its nutritional composition, flavor, and price. Traditional rice traceability detection methods face challenges such as complex detection procedures, high equipment costs, and time inefficiency. This study proposes a rapid detection method for rice origin based on an electronic tongue (ET) combined with a meta-learning and contrastive learning dual aggregation network (MCDANet). Firstly, the taste response signals of rice samples from different origins are collected by an ET device. A dual aggregation model is proposed to analyze the collected ET signals. This model employs a prototypical network based on meta-learning as the inner model for feature extraction and pattern recognition of ET signals. To enhance the feature extraction capability of the inner model, a SimSiam network based on contrastive learning is incorporated as the outer model to optimize the prototypical network. Finally, the proposed model is trained with a semi-supervised strategy under few sample conditions. The experimental results demonstrate that this method achieves superior recognition accuracy in classifying ET signals of different rice samples, with an overall accuracy of 98%. This study provides a novel and rapid detection method for rice origin traceability, which has a promising application in agriculture and the food industry.