Scholarlink

<p>Wines from different geographical origins can exhibit similar visual and aromatic characteristics, while their flavor profiles and market value differ significantly. Traditional identification methods are frequently hindered by complex procedures and high analytical costs. This study proposes a rapid and efficient method for wine origin identification that achieves high accuracy with low computational costs. Specifically, the taste, olfactory, and visual profiles of wine samples are captured using an electronic tongue (ET), electronic nose (EN), and electronic eye (EE), respectively. A MobileNet-Mamba hybrid model is developed to extract both global and local features from the multi-sensory data. Subsequently, a hybrid attention fusion module is proposed. It combines bidirectional and self-attention mechanisms to effectively fuse multi-scale multimodal features. The fused features are then input into the decision network for final origin classification. To ensure optimal performance, Bayesian optimization is employed to fine-tune the model hyperparameters. Experimental results demonstrate the proposed model achieves an accuracy, precision, recall, and F1-score of 99.35%, 99.37%, 99.41%, and 99.37%, respectively. The proposed model has only 6.64&#xa0;M parameters, rendering it suitable for deployment on low-cost edge computing devices. This study provides a novel method for the rapid and on-site identification of wine origin, which has great potential for applications in traceability detection of other beverage types.</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

Synergistic application of electronic tongue, nose, and eye with a lightweight MobileNet-Mamba hybrid model for rapid wine origin identification

  • 期刊论文

  • 发表日期：

    2026年7月13日

查看全文
  • Jingbao Wang,
  • Zhiqiang Wang,
  • Jinyang Zhang,
  • Junying Dou,
  • Shanhui Han,
  • Yubin Lan

摘要

Wines from different geographical origins can exhibit similar visual and aromatic characteristics, while their flavor profiles and market value differ significantly. Traditional identification methods are frequently hindered by complex procedures and high analytical costs. This study proposes a rapid and efficient method for wine origin identification that achieves high accuracy with low computational costs. Specifically, the taste, olfactory, and visual profiles of wine samples are captured using an electronic tongue (ET), electronic nose (EN), and electronic eye (EE), respectively. A MobileNet-Mamba hybrid model is developed to extract both global and local features from the multi-sensory data. Subsequently, a hybrid attention fusion module is proposed. It combines bidirectional and self-attention mechanisms to effectively fuse multi-scale multimodal features. The fused features are then input into the decision network for final origin classification. To ensure optimal performance, Bayesian optimization is employed to fine-tune the model hyperparameters. Experimental results demonstrate the proposed model achieves an accuracy, precision, recall, and F1-score of 99.35%, 99.37%, 99.41%, and 99.37%, respectively. The proposed model has only 6.64 M parameters, rendering it suitable for deployment on low-cost edge computing devices. This study provides a novel method for the rapid and on-site identification of wine origin, which has great potential for applications in traceability detection of other beverage types.