Design of a Spectral Signature Database for Fruit Adulteration Detection Using Machine Learning

Abstract

The detrimental consequences of artificial fruit ripening on human health and food quality have made food adulteration a serious concern. This study suggests a hyperspectral spectral database framework that uses machine learning and spectral signature analysis to identify fruits that have been chemically and naturally ripened. The ASD Field Spec® 4 spectroradiometer was used to gather the spectral signature of banana, mango, and apple samples in the 350–2500 nm wavelength range. Under both natural and chemically ripened circumstances, 1430 spectral signatures were obtained from several fruit varietals. For effective storage, preprocessing, retrieval, and spectral library creation, the gathered hyperspectral data were processed using RS3 software and arranged into a structured Fruit Database. To increase feature extraction and spectrum quality, preprocessing methods including normalisation and second derivative analysis were used. Reflectance spectra in the Visible (VIS), Near Infrared (NIR), and Short-Wave Infrared (SWIR) regions could be visualised and compared thanks to the created spectral library. The suggested framework offers a dependable, non-destructive, and scalable method for intelligent fruit adulteration detection and food quality monitoring applications. It also facilitates the effective management of high-dimensional hyperspectral data.

Introduction

This study focuses on detecting fruit adulteration caused by artificial ripening chemicals, particularly calcium carbide, using hyperspectral analysis, spectral signature databases, and machine learning techniques. Artificial ripening is widely used to accelerate fruit maturation and meet market demand, but chemicals such as calcium carbide release acetylene gas containing harmful contaminants like arsenic and phosphorus, which can cause serious health problems including neurological disorders, respiratory diseases, and cancer. Therefore, distinguishing naturally ripened fruits from chemically ripened ones is an important challenge in food safety and quality assurance.

The research proposes a structured spectral signature database system that integrates hyperspectral data analysis, database management, and machine learning to identify fruit adulteration accurately. Hyperspectral imaging captures detailed reflectance information across a wide range of wavelengths, producing unique spectral fingerprints that reveal chemical and physical changes in fruits that are invisible to the human eye. These spectral signatures can be analyzed to detect the effects of artificial ripening and food adulteration.

The literature review highlights the successful application of machine learning techniques such as Support Vector Machines (SVM), Random Forest (RF), Partial Least Squares Discriminant Analysis (PLS-DA), Artificial Neural Networks (ANN), and Convolutional Neural Networks (CNNs) in fruit classification and adulteration detection. Researchers have also demonstrated the importance of preprocessing methods such as normalization, smoothing, and derivative spectroscopy, along with dimensionality reduction techniques like PCA, t-SNE, and UMAP, to improve classification performance. However, previous studies often lacked structured database frameworks, scalability, or real-world validation.

The proposed framework begins with collecting spectral signatures using the ASD FieldSpec® 4 Spectroradiometer, which operates across the 350–2500 nm wavelength range covering the Visible (VIS), Near Infrared (NIR), and Short-Wave Infrared (SWIR) regions. Spectral data are acquired from naturally and chemically ripened bananas, mangoes, and apples. The collected data are processed using specialized software, organized into a spectral library, stored in a structured database, and then analyzed for adulteration detection.

A total of 1,430 spectral signatures were collected and stored in the database. The dataset includes 360 banana spectra, 660 mango spectra, and 410 apple spectra, representing both naturally and chemically ripened samples. Various fruit varieties were included to improve spectral diversity and model generalization. The database stores raw spectral data, preprocessing outputs, extracted features, dimensionality reduction results, and classification labels, enabling efficient retrieval, visualization, machine learning training, and future scalability.

The study concludes that combining hyperspectral imaging, structured spectral databases, preprocessing techniques, and machine learning algorithms provides an effective, scalable, and non-destructive approach for detecting fruit adulteration. The developed framework enhances food safety monitoring by enabling accurate identification of chemically ripened fruits and establishing a standardized platform for future intelligent food quality assessment systems.

Conclusion

This study suggested a hyperspectral spectral database system that uses machine learning and spectral signature analysis to identify both chemically and naturally ripened crops. The ASD Field Spec® 4 spectroradiometer was used to gather spectral data of banana, mango, and apple samples in the 350–2500 nm wavelength range. Hyperspectral data could be efficiently stored, pre-processed, visualised, and analysed thanks to the developed Fruit Database and spectral library system. For intelligent fruit adulteration detection and food quality monitoring applications, the suggested framework offers a dependable, non-destructive, and scalable method.

References

[1] Y. Huang, R. Lu, and K. Chen, “Detection of fruit ripeness using hyperspectral imaging and machine learning,” Journal of Food Engineering, vol. 256, pp. 1–10, 2022. [2] J. Qin, K. Chao, and M. Kim, “Hyperspectral imaging technology for food and agricultural products,” Food Engineering Reviews, vol. 5, no. 3, pp. 157–171, 2021. [3] S. Siddiqui and M. Dhua, “Eating artificially ripened fruits is harmful,” Current Science, vol. 99, no. 12, pp. 1664–1668, 2010. [4] World Health Organization, “Calcium carbide and food safety concerns,” WHO Food Safety Reports, 2020. [5] A. A. Gowen, C. P. O’Donnell, P. J. Cullen, G. Downey, and J. M. Frias, “Hyperspectral imaging – an emerging process analytical tool for food quality and safety control,” Trends in Food Science & Technology, vol. 18, no. 12, pp. 590–598, 2021. [6] D. Wu, D. W. Sun, and Y. He, “Application of hyperspectral imaging for food quality analysis and adulteration detection,” Critical Reviews in Food Science and Nutrition, vol. 59, no. 1, pp. 1–15, 2023. [7] A. Mishra, S. Sharma, and P. Verma, “Deep learning approaches for banana quality assessment using image analysis,” Computers and Electronics in Agriculture, vol. 185, pp. 106139, 2021. [8] M. ElMasry, N. Wang, and C. Vigneault, “Detecting apple bruises using hyperspectral imaging,” Computers and Electronics in Agriculture, vol. 71, pp. S27–S35, 2020. [9] L. Sun, X. Wang, and H. Zhang, “Deep learning-based hyperspectral imaging for agricultural product classification,” IEEE Access, vol. 11, pp. 22456–22470, 2024. [10] R. Lu, D. Wu, and Y. He, “Spectral database systems for food quality monitoring,” Journal of Food Measurement and Characterization, vol. 15, no. 4, pp. 2100–2112, 2022. [11] P. Geladi and H. Grahn, Hyperspectral Image Analysis, Wiley Publications, 2021. [12] Y. Huang, R. Lu, and K. Chen, “Detection of fruit ripeness using hyperspectral imaging and machine learning,” Journal of Food Engineering, vol. 256, pp. 1–10, 2022. [13] A. Mishra, S. Sharma, and P. Verma, “Deep learning approaches for banana quality assessment using image analysis,” Computers and Electronics in Agriculture, vol. 185, pp. 106139, 2021. [14] J. Qin, K. Chao, and M. Kim, “Hyperspectral imaging technology for food and agricultural products,” Food Engineering Reviews, vol. 5, no. 3, pp. 157–171, 2021. [15] R. Lu, D. Wu, and Y. He, “Spectral database systems for food quality monitoring,” Journal of Food Measurement and Characterization, vol. 15, no. 4, pp. 2100–2112, 2022. [16] A. A. Gowen, C. P. O’Donnell, P. J. Cullen, G. Downey, and J. M. Frias, “Hyperspectral imaging – an emerging process analytical tool for food quality and safety control,” Trends in Food Science & Technology, vol. 18, no. 12, pp. 590–598, 2021. [17] D. Wu, D. W. Sun, and Y. He, “Machine learning and hyperspectral imaging for chemically ripened fruit detection,” IEEE Access, vol. 11, pp. 22456–22470, 2024. [18] M. ElMasry, N. Wang, and C. Vigneault, “Detecting apple bruises using hyperspectral imaging,” Computers and Electronics in Agriculture, vol. 71, pp. S27–S35, 2020. [19] L. Sun, X. Wang, and H. Zhang, “Deep learning-based hyperspectral imaging for agricultural product classification,” IEEE Access, vol. 11, pp. 22456–22470, 2024. [20] Y. He, F. Liu, and D. Wu, “Deep learning applications in hyperspectral food analysis,” Food Analytical Methods, vol. 16, no. 5, pp. 1123–1138, 2023. [21] P. Geladi and H. Grahn, Hyperspectral Image Analysis, Wiley Publications, 2021. [22] T. Chen, Y. Zhang, and H. Li, “Random forest based spectroscopy analysis for food adulteration detection,” Sensors, vol. 22, no. 14, pp. 5400–5415, 2022. [23] J. Workman and L. Weyer, Practical Guide to Interpretive Near-Infrared Spectroscopy, CRC Press, 2021. [24] Food and Agriculture Organization, “Food safety and quality monitoring using advanced sensing technologies,” FAO Technical Report, 2023.

Copyright

Copyright © 2026 Snehal S. Datwase, Dr. Charansing N. Kayte, Dr. R. R. Deshmukh. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.