Scholarlink

<p>Durian maturity critically affects consumer acceptance, market value, and postharvest losses, yet its assessment remains largely subjective. This study presents a non-destructive multimodal sensing system integrating gas/VOC sensing, thermal imaging, and acoustic measurement for durian maturity classification. A dataset of 90 fruits representing three ripeness stages, namely unmature, partially mature, and mature, was evaluated using supervised machine learning with a fruit-level grouped validation procedure to reduce potential data leakage from repeated measurements. Gas/VOC sensors captured maturity-related changes in alcohol, VOC-related response, CO₂, and O₂, while thermal imaging provided surface temperature information and acoustic measurement provided tapping-based dB response. These sensor responses were consistent with destructive validation trends, including decreasing flesh firmness and increasing soluble solids and alcohol content. Among the evaluated models, the neural network achieved the highest performance, with 96.91% accuracy and an AUC of 0.98. Ablation analysis showed that gas/VOC features were the main contributors to classification performance, thermal features provided complementary information, and scalar acoustic dB alone had limited discriminatory ability. These findings demonstrate the potential of multimodal sensor fusion and machine learning for non-destructive durian maturity assessment.</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

Development of non-destructive durian fruit maturity detection tool based on multi-variable sensor for harvest quality optimisation

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月11日

查看全文
  • Aulia Brilliantina,
  • Tri Agus Siswoyo,
  • Yuli Witono,
  • Bayu Taruna Widjaja Putra

摘要

Durian maturity critically affects consumer acceptance, market value, and postharvest losses, yet its assessment remains largely subjective. This study presents a non-destructive multimodal sensing system integrating gas/VOC sensing, thermal imaging, and acoustic measurement for durian maturity classification. A dataset of 90 fruits representing three ripeness stages, namely unmature, partially mature, and mature, was evaluated using supervised machine learning with a fruit-level grouped validation procedure to reduce potential data leakage from repeated measurements. Gas/VOC sensors captured maturity-related changes in alcohol, VOC-related response, CO₂, and O₂, while thermal imaging provided surface temperature information and acoustic measurement provided tapping-based dB response. These sensor responses were consistent with destructive validation trends, including decreasing flesh firmness and increasing soluble solids and alcohol content. Among the evaluated models, the neural network achieved the highest performance, with 96.91% accuracy and an AUC of 0.98. Ablation analysis showed that gas/VOC features were the main contributors to classification performance, thermal features provided complementary information, and scalar acoustic dB alone had limited discriminatory ability. These findings demonstrate the potential of multimodal sensor fusion and machine learning for non-destructive durian maturity assessment.