Combining image processing technique and three artificial intelligence methods to recognize the freshness of freshwater shrimp

Abstract

Since seafood is highly susceptible to corruption, it is important to check their storage and shelf-life time. In this research, image processing technology was used to recognize the freshness (time lasted of catching) of shrimps. Shrimp samples were randomly selected from shrimp farming pools and stored in three storage conditions: freezer, refrigerator, and cool environments. Images were taken from the samples at intervals of two hours in a controlled environment for more than a month. Finally, 482 properties were extracted from each image. Three effective parameters for modeling were selected by sensitivity analysis. The time that lasted from catching was the output of the models. Modeling was performed using ANFIS, ANN, and RSM algorithms. In the modeling, the <em>R</em>² values of the ANN algorithm with 0.987006, 0.987009, 0.984484, and 0.976001 were the best model for storing conditions: freezer, refrigerator, cooler environments and the total of storage conditions, respectively. All three modeling methods can estimate the catching time with high accuracy. But the ANN model was recognized as the best one according to the remaining diagram and the values of R² and MSE.