Scholarlink

<p>The geographic origin of azuki beans plays a key role, ensuring product quality and maintaining market order. In this work, a novel spectral feature calculation and classification network (SFCC-Net), integrated with a hyperspectral system, was proposed to achieve rapid, non-destructive, and accurate identification of azuki bean origins. First, using a hyperspectral system, the reflectance spectral data of azuki bean samples were collected within the 900–1700&#xa0;nm wavelength range, covering six major producing origins in China. To reduce noise interference and streamline the data, the average spectrum of the region of interest (ROI) was extracted. Second, a spectral feature calculation module (SFCM) was designed. It incorporates convolutional mechanisms, spatial attention, channel attention, and self-attention mechanisms. This module first performed local feature extraction, then conducted multidimensional optimization, and finally realized global correlation fusion, automatically uncovering deep discriminative features. Finally, an end-to-end SFCC-Net classification network was constructed based on the SFCM. Through structural optimization, ablation studies, and performance comparison with state-of-the-art spectral data classification methods, SFCC-Net achieved an accuracy of 98.00%, a precision of 98.11%, a recall of 97.89%, and an F1-score of 98.10%, while also attaining the optimal classification stability. This study provides an efficient and reliable technical solution for identifying the origin of azuki beans, and offers a reference for the design of deep learning models in the field of spectral detection of agricultural products.</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

Origin identification of azuki beans: a novel spectral feature computation and classification network integrated with a hyperspectral system

  • 期刊论文

  • 发表日期：

    2026年6月24日

查看全文
  • Qiufen Wang,
  • Tianhao Wang,
  • Lu An

摘要

The geographic origin of azuki beans plays a key role, ensuring product quality and maintaining market order. In this work, a novel spectral feature calculation and classification network (SFCC-Net), integrated with a hyperspectral system, was proposed to achieve rapid, non-destructive, and accurate identification of azuki bean origins. First, using a hyperspectral system, the reflectance spectral data of azuki bean samples were collected within the 900–1700 nm wavelength range, covering six major producing origins in China. To reduce noise interference and streamline the data, the average spectrum of the region of interest (ROI) was extracted. Second, a spectral feature calculation module (SFCM) was designed. It incorporates convolutional mechanisms, spatial attention, channel attention, and self-attention mechanisms. This module first performed local feature extraction, then conducted multidimensional optimization, and finally realized global correlation fusion, automatically uncovering deep discriminative features. Finally, an end-to-end SFCC-Net classification network was constructed based on the SFCM. Through structural optimization, ablation studies, and performance comparison with state-of-the-art spectral data classification methods, SFCC-Net achieved an accuracy of 98.00%, a precision of 98.11%, a recall of 97.89%, and an F1-score of 98.10%, while also attaining the optimal classification stability. This study provides an efficient and reliable technical solution for identifying the origin of azuki beans, and offers a reference for the design of deep learning models in the field of spectral detection of agricultural products.