Scholarlink

<p>Addressing the demand for rapid, non-destructive meat quality assessment, this study developed a Vision-Based Pork Freshness Detection System (VPFDS) that integrates an anthocyanin-based colorimetric hydrogel sensor with machine vision and deep learning. The hydrogel sensor exhibited distinct pH and ammonia-responsive color transitions associated with the accumulation of volatile basic nitrogen compounds during pork spoilage. Among the investigated formulations, GE/ZnO/MA2 showed the most pronounced color change, making it optimal for freshness monitoring. A dedicated machine vision platform with controlled illumination was constructed to ensure stable image acquisition and eliminate variations commonly encountered in smartphone-based detection. Four convolutional neural network (CNN) models were evaluated for automated ternary freshness classification (“fresh”, “less fresh”, and “spoiled”). EfficientNetB0 achieved the best performance, with an overall accuracy of 94%, precision values above 0.90 for all freshness categories, and F1-scores of 0.92, 0.91, and 0.99 for fresh, less-fresh, and spoiled pork, respectively. Rigorous validation under real-world conditions confirmed the system’s reliability: results delivered within 2&#xa0;s show high agreement with standard TVB-N measurements. This work provides a practical, rapid alternative to traditional destructive methods, offering significant potential for intelligent meat quality control in industrial applications.</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

Real-Time, Non-destructive Pork Freshness Monitoring: Automated Ternary Grading via Anthocyanin-Based Smart Sensor and Machine Vision

  • 期刊论文

  • 发表日期：

    2026年7月31日

查看全文
  • Zhiran Liang,
  • Jingxuan Qiu,
  • Caiyun Liu,
  • Shoaib Younas,
  • Xin Wang

摘要

Addressing the demand for rapid, non-destructive meat quality assessment, this study developed a Vision-Based Pork Freshness Detection System (VPFDS) that integrates an anthocyanin-based colorimetric hydrogel sensor with machine vision and deep learning. The hydrogel sensor exhibited distinct pH and ammonia-responsive color transitions associated with the accumulation of volatile basic nitrogen compounds during pork spoilage. Among the investigated formulations, GE/ZnO/MA2 showed the most pronounced color change, making it optimal for freshness monitoring. A dedicated machine vision platform with controlled illumination was constructed to ensure stable image acquisition and eliminate variations commonly encountered in smartphone-based detection. Four convolutional neural network (CNN) models were evaluated for automated ternary freshness classification (“fresh”, “less fresh”, and “spoiled”). EfficientNetB0 achieved the best performance, with an overall accuracy of 94%, precision values above 0.90 for all freshness categories, and F1-scores of 0.92, 0.91, and 0.99 for fresh, less-fresh, and spoiled pork, respectively. Rigorous validation under real-world conditions confirmed the system’s reliability: results delivered within 2 s show high agreement with standard TVB-N measurements. This work provides a practical, rapid alternative to traditional destructive methods, offering significant potential for intelligent meat quality control in industrial applications.