Scholarlink

<p>The assurance of food safety and quality, particularly in fresh produce, is a critical concern in modern food systems. Gamma irradiation has emerged as an effective method for extending shelf life and reducing microbial contamination in food without causing observable changes to its physical appearance. However, the absence of visual indicators post-irradiation necessitates the development of reliable, non-destructive detection techniques to verify exposure and dosage. Conventional detection methods, including chemical and physical assays, are often invasive, destructive, and impractical for routine screening. This study presents a novel, non-destructive framework for identifying irradiated food using hyperspectral imaging (HSI) integrated with advanced deep learning networks. The proposed system captures hyperspectral data from apple samples before and after gamma irradiation, followed by spectral feature extraction using endmember detection algorithms such as the Pixel Purity Index (PPI) and Fast Iterative Pixel Purity Index (FIPPI). These features are then used to train and evaluate deep learning classifiers, including Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) networks. Experimental results demonstrate that the combination of PPI based on Principal Component Analysis (PCA) as a reduction method and LSTM and BILSTM achieves a classification accuracy of 94%, effectively distinguishing irradiated samples from non-irradiated counterparts. This approach demonstrates the effectiveness of hyperspectral imaging combined with deep learning as a nondestructive, accurate, and practical solution for detecting irradiation food. It offers significant potential for enhancing quality assurance in food safety monitoring. This approach ensures the integrity of the client’s samples by eliminating the need for destructive chemical or physical analyses, thereby preserving the sample while reliably verifying radiation exposure.</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 irradiated food through hyperspectral imaging assisted by deep learning techniques

  • 期刊论文

  • 发表日期：

    2025年10月1日

查看全文
  • H. M. Nada,
  • Osama A. Omer,
  • Hamada A. H. Esmaiel,
  • A. A. Arafa,
  • M. Ashour

摘要

The assurance of food safety and quality, particularly in fresh produce, is a critical concern in modern food systems. Gamma irradiation has emerged as an effective method for extending shelf life and reducing microbial contamination in food without causing observable changes to its physical appearance. However, the absence of visual indicators post-irradiation necessitates the development of reliable, non-destructive detection techniques to verify exposure and dosage. Conventional detection methods, including chemical and physical assays, are often invasive, destructive, and impractical for routine screening. This study presents a novel, non-destructive framework for identifying irradiated food using hyperspectral imaging (HSI) integrated with advanced deep learning networks. The proposed system captures hyperspectral data from apple samples before and after gamma irradiation, followed by spectral feature extraction using endmember detection algorithms such as the Pixel Purity Index (PPI) and Fast Iterative Pixel Purity Index (FIPPI). These features are then used to train and evaluate deep learning classifiers, including Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) networks. Experimental results demonstrate that the combination of PPI based on Principal Component Analysis (PCA) as a reduction method and LSTM and BILSTM achieves a classification accuracy of 94%, effectively distinguishing irradiated samples from non-irradiated counterparts. This approach demonstrates the effectiveness of hyperspectral imaging combined with deep learning as a nondestructive, accurate, and practical solution for detecting irradiation food. It offers significant potential for enhancing quality assurance in food safety monitoring. This approach ensures the integrity of the client’s samples by eliminating the need for destructive chemical or physical analyses, thereby preserving the sample while reliably verifying radiation exposure.