Scholarlink

<p>The adulteration of black tea by mixing old and new tea leaves to pass off inferior products as premium ones severely infringes upon consumer rights. This paper proposes an electronic nose (e-nose) system based on a Dual-branch Gas Feature Classification Network (DGFC-Net) for the rapid and accurate detection of black tea adulteration. Addressing the issues of strong subjectivity and low efficiency in traditional sensory evaluation and physicochemical analysis methods, this study employs the e-nose system to collect volatile organic compound signals from black tea samples with varying adulteration ratios using a 10 metal-oxide-semiconductor (MOS) sensor array. By integrating local feature convolution with a lightweight self-attention mechanism, a Dual-branch Gas Feature Calculation Module (DGFCM) is designed to effectively fuse the cross-sensitive correlations among sensors and temporal dynamic characteristics for computing deep gas features. Furthermore, a lightweight DGFC-Net is developed based on the DGFCM. Experimental results demonstrate that DGFC-Net achieves an accuracy of 98.10%, a precision of 97.92%, and a recall of 98.67% in classifying black tea samples with six adulteration ratios. Its classification performance significantly outperforms classical lightweight convolutional, self-attention, and hybrid networks, as well as state-of-the-art gas information recognition methods. This research provides an efficient and reliable technical solution for non-destructive quality detection of black tea, holding significant application value for food adulteration regulation.</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

Identification of black tea adulteration using a dual-branch gas feature classification network and an electronic nose system

  • 期刊论文

  • 发表日期：

    2026年3月10日

查看全文
  • Ke Chen,
  • Yuetong Si,
  • Liming Liu

摘要

The adulteration of black tea by mixing old and new tea leaves to pass off inferior products as premium ones severely infringes upon consumer rights. This paper proposes an electronic nose (e-nose) system based on a Dual-branch Gas Feature Classification Network (DGFC-Net) for the rapid and accurate detection of black tea adulteration. Addressing the issues of strong subjectivity and low efficiency in traditional sensory evaluation and physicochemical analysis methods, this study employs the e-nose system to collect volatile organic compound signals from black tea samples with varying adulteration ratios using a 10 metal-oxide-semiconductor (MOS) sensor array. By integrating local feature convolution with a lightweight self-attention mechanism, a Dual-branch Gas Feature Calculation Module (DGFCM) is designed to effectively fuse the cross-sensitive correlations among sensors and temporal dynamic characteristics for computing deep gas features. Furthermore, a lightweight DGFC-Net is developed based on the DGFCM. Experimental results demonstrate that DGFC-Net achieves an accuracy of 98.10%, a precision of 97.92%, and a recall of 98.67% in classifying black tea samples with six adulteration ratios. Its classification performance significantly outperforms classical lightweight convolutional, self-attention, and hybrid networks, as well as state-of-the-art gas information recognition methods. This research provides an efficient and reliable technical solution for non-destructive quality detection of black tea, holding significant application value for food adulteration regulation.