Document Type : Review Article

Author

• Seyed Reza Movahhed Ghodsinya

Assistant Professor, Electrical and Computer Department, Technical and Engineering Faculty, Torbat Heydarieh University, Torbat Heydarieh, Iran.

Abstract

Introduction: Saffron, a prized spice with extensive applications in both culinary and medicinal realms, is extracted from the stigmas of the Crocus sativus flower, boasting a distinctive aroma and flavor profile. Its exceptional value renders it susceptible to adulteration, a process involving the addition of less expensive components to increase quantity or reduce production costs. Adulteration poses a substantial risk to quality and purity of saffron, potentially resulting in health concerns and economic setbacks. Therefore, ensuring the genuineness of saffron products is imperative. Machine vision techniques have emerged as a promising solution for detecting saffron adulteration and elevating product quality.
 
Materials and Methods: To assess the current state of research on how machine vision is used in the field of saffron plant, we conducted a comprehensive literature review utilizing various scientific databases, including Scopus, Web of Science, and PubMed. The search was conducted using a combination of keywords such as saffron, image processing, adulteration, and quality control. Articles were screened based on their relevance to the subject matter and adherence to inclusion criteria.
 
Results and Discussion: Our review underscores saffron image processing as a burgeoning research domain with the potential to substantially enhance the quality and purity of saffron products. Diverse imaging techniques have been employed for capturing saffron imagery, including RGB imaging, HSI, and hyperspectral imaging. RGB imaging, a straightforward and widely adopted method, captures images in the red, green, and blue channels. HSI imaging, on the other hand, captures images across hue, saturation, and intensity channels, while hyperspectral imaging records images at multiple wavelengths. Image preprocessing plays a pivotal role in saffron image processing, encompassing noise reduction, color balance correction, and contrast enhancement. Feature extraction and classification are equally crucial steps, involving the identification and selection of pertinent image features and their subsequent categorization as authentic or adulterated. A variety of methodologies have been devised for saffron adulteration detection, including chemometric analysis, machine learning, and deep learning. Chemometric analysis employs statistical techniques to analyze the chemical composition of saffron samples and identify adulterants. Machine learning, a facet of artificial intelligence, entails training computer models on datasets to predict the authenticity of new samples. Deep learning, a more advanced variant, employs artificial neural networks to extract features from the images and classify them. While chemometric analysis has found widespread application in saffron adulteration detection and yielded promising outcomes, recent studies indicate the potential of machine and deep learning. Deep learning models such as convolutional neural networks and recurrent neural networks have been instrumental in feature extraction from saffron images and the subsequent authentication of their purity.
 
Conclusion: To conclude, our review underscores the critical role of machine vision in safeguarding the quality and purity of saffron products. The application of diverse imaging techniques and detection methodologies has demonstrated remarkable promise in detecting saffron adulteration. Nevertheless, further research is imperative to refine and advance the accuracy and reliability of saffron image processing techniques, particularly within the domain of saffron adulteration detection. Given the escalating demand for high-quality saffron products, the development of effective saffron image processing techniques stands as a critical factor in ensuring consumer trust and safety.

Keywords

• Image processing
• Machine vision
• Saffron
• Stigma
• Grading

20.1001.1.23453869.1402.11.1.12.3