Scholarlink

<p>To achieve efficient and rapid detection of durian sugar content—with potential extension to maturity assessment of other fruits—this study proposes a novel hybrid framework integrating Fast Continuous Wavelet Transform (FCWT) and Spatiotemporal Backpropagation-based Spiking Neural Network (STBP-SNN). Hyperspectral images of durian pulp were collected within the 900–1700&#xa0;nm wavelength range, and samples were categorized into three sugar levels (high, medium, low) to establish a targeted evaluation system for sugar content grading.​FCWT was employed for multi-scale time-frequency analysis, converting one-dimensional spectral curves into two-dimensional feature matrices to capture fine-grained spectral variations that are critical for distinguishing sugar levels. These feature matrices were then input to the STBP-SNN, which leverages the unique dynamics of spiking neurons to effectively extract spatiotemporal features from hyperspectral data and perform accurate classification.​ The experimental results indicated that the proposed framework achieved an average test accuracy of around 96%, highlighting its robustness in discriminating durian sugar content. By integrating advanced spectral feature extraction (FCWT) with intelligent spatiotemporal classification (STBP-SNN), the framework provides a scalable and reliable solution for fruit quality assessment, especially for precision grading of internal quality indicators like sugar content.</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

Efficient durian sugar content grading via hyperspectral imaging and fast continuous wavelet transform and spiking neural network framework

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月25日

查看全文
  • Xing Qin,
  • Xiaoheng Zhang,
  • Chenxiao Lai,
  • Hanliang Liang,
  • Liyu Li,
  • Chu Qin

摘要

To achieve efficient and rapid detection of durian sugar content—with potential extension to maturity assessment of other fruits—this study proposes a novel hybrid framework integrating Fast Continuous Wavelet Transform (FCWT) and Spatiotemporal Backpropagation-based Spiking Neural Network (STBP-SNN). Hyperspectral images of durian pulp were collected within the 900–1700 nm wavelength range, and samples were categorized into three sugar levels (high, medium, low) to establish a targeted evaluation system for sugar content grading.​FCWT was employed for multi-scale time-frequency analysis, converting one-dimensional spectral curves into two-dimensional feature matrices to capture fine-grained spectral variations that are critical for distinguishing sugar levels. These feature matrices were then input to the STBP-SNN, which leverages the unique dynamics of spiking neurons to effectively extract spatiotemporal features from hyperspectral data and perform accurate classification.​ The experimental results indicated that the proposed framework achieved an average test accuracy of around 96%, highlighting its robustness in discriminating durian sugar content. By integrating advanced spectral feature extraction (FCWT) with intelligent spatiotemporal classification (STBP-SNN), the framework provides a scalable and reliable solution for fruit quality assessment, especially for precision grading of internal quality indicators like sugar content.