An Optimal Faster Region-Based Convolutional Neural Network for Oil Adulteration Detection

• Published in: Arabian journal for science and engineering (2011) Vol. 48; no. 2; pp. 2231 - 2245
• Main Authors: Surya, V., Senthilselvi, A.
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.02.2023; Springer Nature B.V
• Subjects: Artificial neural networks; Engineering; Humanities and Social Sciences; multidisciplinary; Neural networks; Rapeseed; Research Article-Computer Engineering and Computer Science; Science; Vegetable oils; Oil adulteration; War strategy optimization algorithm; Chemometric methods; Excitation-emission matrix fluorescence; Faster region-based convolutional neural network
• ISSN: 2193-567X; 1319-8025; 2191-4281
• DOI: 10.1007/s13369-022-07115-7

Natural oils such as avocado oil (AO), corn oil (CO), chamomile oil (CMO), and rapeseed oil (RSO) are used for improving health, skin, and hair care for centuries in different parts of the world. The AO, CO, and RSO oil are known for their antibacterial and moisturizing properties. The growth of the modern health and wellness industries has deceived people to find affordable, additive-free, and effective products. Consumers are always willing to pay a higher price for the unadulterated oils due to the major benefits it offers in terms of complexion and health. However, the AO, CMO, and RSO oils are rarely available in the market and the ones available are subject to adulteration risks which affect the consumer’s health and also violate their rights. This paper presents a war strategy optimized faster Region-based Convolutional Neural Network (RCNN) architecture named Modified Faster RCNN for oil adulteration detection. The pure natural AO, CO, CMO, and RSO oils are adultered with different oils such as vegetable oils with different ratios and these oils themselves. Spectra analysis was conducted to identify the adultered vegetable oil samples and the Modified Faster RCNN was employed to classify the pure oil from these unadultered samples. The results show that the proposed model is effective in rapidly identifying the adulterated components presented in high-quality oils.