Application of fuzzy logic and neural-fuzzy inference system (ANFIS) for prediction of physicochemical changes and quality classification of coated sweet lemon during storage

The base of intelligent methods is using hidden knowledge in the experimental data, trying to extract the inherent relationships among them and generalizing results to other situations. Artificial neural networks are one of the most essential methods used in the field of artificial intelligence was inspired by how the human brain works, training takes place first, and then the information related to the data is stored in the form of the network's weights. Fuzzy logic is an important decision-making tool that has recently found some applications in food quality. Also, it is possible to find out the reasons for low and high ranking of products evaluated by the judges. The fuzzy model can be used to determine the importance of individual factors to the overall quality of a product. The ANFIS model is a combination of the artificial neural network (ANN) and a fuzzy inference system (FIS) in such a way that the neural networks are applied to determine the parameter of the fuzzy inference system. The fuzzy logic theory effectively addresses the uncertainty problems that solve the ambiguity. Sweet lemon (Citrus limetta) fruit is a popular agricultural product cultivated in tropical countries used to treat common colds, influenza and hypertension. Sweet lemon is quite perishable with postharvest losses such as weight loss, physiological deterioration, decay, and softening texture. The objective of the present study was to investigate grading of sweet lemon fruit based on quality and visual characteristics using fuzzy logic and ANFIS.  

Ripe sweet lemon (Citrus limetta) fruits and radish (Raphanus sativus L.) leaves were purchased from the local market in Zanjan, Iran. For emulsion solution preparation, 50 ml alginate sodium solution, 1 ml glycerin and 0 or 10 g radish leaf extract were mixed, then the coating solution volume was made up to 100 ml using distilled water. Finally, the mixture was steered for 200 second. Sweet lemon fruits were dipped in coating solutions or distillate water (for control treatment) for 2 minutes at ambient temperature (25℃) and were then air-dried for 2 h using a fan. All treatments stored at 4℃ for 50 days. Firmness, pH, titratable acidity (TA), total soluble solids content (TSS), color, and shape were measured at 10-day intervals. This paper introduces an adaptive neural-fuzzy inference system (ANFIS) model to classify sweet lemon based on the quality parameters and RGB intensity values. The ANFIS with different types of input membership functions (MFs) was developed. A study was performed using fuzzy logic and adaptive neural-fuzzy inference system (ANFIS) to predict the quality parameters of sweet lemon (firmness and ripening index).  

Our results showed that ‘triangle2mf’ MF performs much better than other mentioned MFs for defect inspection. The classification accuracy of the ANFIS with ‘triangle2mf’ MF was 97.5% and 96.6% for quality input and visual input, respectively, and the total correct classification rate was 97.01%. Therefore, this study indicated the possibility of developing a potentially useful classification tool using the ANFIS technique based on quality parameters and RGB values for fruit classification during processing, storage and distribution. Comparing the results obtained from fuzzy logic with various membership function, showed that the RMSE in the fuzzy logic with ‘guss2mf’ MF was lower than other algorithms. The proposed approach focuses on three research motivations. First, to develop a fuzzy rule-based classification system that can detect all the four quality grades of the sweet lemon. Second, the system should be able to predict the quality parameters of sweet lemon. Fuzzy logic deserved high level of accuracy in classification of sweet lemon, indicating high correlation between the data obtained from Mamdani fuzzy rules and experimental ones during storage time.