Distributional uniformity quantification in heterogeneous prepared dishes combined the hyperspectral imaging technology with Moran’s I: A case study of pizza

• Published in: Food chemistry 2025-02, Vol.466, p.141511, Article 141511
• Main Authors: Gao, Peipei, Li, Wenlong, Hashim, Sulafa B.H., Liang, Jing, Xu, Jialong, Huang, Xiaowei, Zou, Xiaobo, Shi, Jiyong
• Format: Article
• Language: English
• Subjects: Distributional uniformity quantification; Food Handling; GLCM; Hyperspectral Imaging - methods; Hyperspectral imaging technology; Image Processing, Computer-Assisted - methods; Machine learning; Moran's I; Prepared dishes; Principal Component Analysis; Quality Control; Support Vector Machine
• ISSN: 0308-8146; 1873-7072; 1873-7072
• DOI: 10.1016/j.foodchem.2024.141511

Quality detection is critical in the development of prepared dishes, with distributional uniformity playing a significant role. This study used hyperspectral imaging (HSI) and Moran's I to quantify distributional uniformity, employing pizza as case. Pizza ingredients' spectra were collected, pre-processed with Detrended Fluctuation Analysis (DFA), Savitzky-Golay (SG) and Standard Normal Variate (SNV), and down-scaled with Principal Component Analysis (PCA). Subsequently, the classifiers Fine Tree, Support Vector Machine (SVM), and K-Nearest Neighbors (KNN) were utilized, where KNN based on the DFA-processed data had the greatest accuracy of 99.2 %. This best-fit model was used to create visualization maps. At last, image analysis methods containing regional statistics, Grey Level Co-occurrence Matrix (GLCM) and Moran's I were used to measure distributional uniformity. Moran's I demonstrated great distinctiveness and accuracy, making it the best tool. Therefore, HSI and Moran's I combination proved feasible to indicate distributional uniformity, ensuring the high quality of prepared dishes. •HSI and Moran's I were combined to quantify uniformity in prepared dishes.•Input data after DFA preprocess and PCA downscale performed better.•Optimal results were obtained from KNN based on DFA-processed data.•Uniformity indicators regional statistics, GLCM and Moran's I were compared.