Loading...

Development and Application of Chemometric Methods for Hyperspectral Image Analysis for Authentication and Adulteration Detection in Food (Saffron and Turmeric)

Hashemi Nasab, Fatemeh Sadat | 2023

128 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: Farsi
  3. Document No: 56124 (03)
  4. University: Sharif University of Technology
  5. Department: Chemistry
  6. Advisor(s): Parastar Shahri, Hadi; Abdollahi, Hamid
  7. Abstract:
  8. The use of hyperspectral images to detect food fraud has become popular and it is necessary to develop chemometrics methods for analyzing the data from these images. Additionally, food authenticity has become a major challenge, and the focus of this thesis is on developing multivariate methods in chemometrics to extract useful information from data obtained from food authenticity verification using hyperspectral imaging (HSI). This thesis consists of six chapters. In the first chapter, a brief introduction to the fundamentals of hyperspectral imaging and food authenticity verification is presented. In the second chapter, the data structure of these images and chemometric methods including preprocessing methods and multivariate analyses are discussed. The following chapters are dedicated to the theoretical development of methods for hyperspectral image analysis and the application of these methods in food authenticity verification. In chapters three and four, multivariate analysis methods, especially independent component analysis (ICA), are discussed. In chapter three, a new approach for examining the independence of feasible solutions in binary systems using mutual information (MI) is presented. Additionally, the concept of the MI map is explained for the first time. Investigating the independence of different solutions can help in a more accurate examination of ICA and multivariate curve resolution (MCR) analysis methods. In the experimental section, various datasets are simulated with different noise levels and degrees of overlap for binary and ternary systems, and feasible solutions are calculated using grid search (GS) and Louton-Silvester (LS) algorithms. The MI map is used to estimate the level of independence between different responses, and the performance of three algorithms, ICA, MF-ICA, and MILCA, is evaluated. The results showed that MF-ICA responses are meaningful within the area of feasible solutions and the pursuit of more independent solutions in this area can lead to a more accurate analysis using MF-ICA. In the fifth and sixth chapters, hyperspectral images are examined in multivariate analysis for the authenticity assessment of two spices, saffron and turmeric. Multivariate classification techniques for authenticity assessment of saffron and detection of fraud are proposed using hyperspectral imaging and the MF-ICA algorithm. 38 authentic Iranian saffron samples were examined using the MF-ICA algorithm, and five fraudulent factors were also investigated. PLS-DA and DD-SIMCA methods were used for supervised classification, and the classification accuracy for most models was 100% for calibration and prediction sets. In chapter six, a rapid and non-destructive analytical method was developed using Vis-SWNIR HSI and chemometric techniques for the authenticity assessment of turmeric powder and identification of five adulterant which are plant-base. Various multivariate separation methods such as MF-ICA and MCR-ALS were used to extract pure component patterns. Then, PCA and DD-SIMCA were used for unsupervised analysis and PLS-DA for the detection of authentic and fraudulent samples. Based on this, modeling results with a sensitivity of 95% and specificity of 100% were obtained. The accuracy of PLS-DA was also 100% for all samples, confirming the validity of the proposed method. The performance of two hyperspectral imaging devices, HYSPIM and SPECIM, was compared for adulteration detection in turmeric
  9. Keywords:
  10. Saffron ; Chemometrics Method ; Multivariate Curve Resolution ; Independent Component Analysis (ICA) ; Rotational Ambiguity ; Adulteration in Foods ; Food Authenticity ; Hyperspectral Imaging ; Feasible Solutions Area

 Digital Object List

 Bookmark

No TOC