Fish freshness categorization from eyes and gills color features using multi-class artificial neural network and support vector machines

Developing new techniques to determine fish freshness and quality can enhance nutritional value of the overall household food basket. In this research, digital image analysis was utilized to assess the freshness of rainbow trout fish by tracing the color attributes of its eyes and gills. The image data were collected from left and right eyes and gills in a 10-day ice-storage duration, and color components were extracted in RGB, HSV, and L*a*b* color spaces. Analysis of variance revealed that the RGB components of both eyes and gills had a significant change towards getting brighter during the ice-storage. Feature extraction was fulfilled from the color spaces, and then artificial neural networks (ANNs) and support vector machines (SVMs) were applied for classification of the ice-storage durations. The overall accuracies of the developed models demonstrated that the ANN somewhat outperformed the SVM for both the extracted features from the eyes and gills. Moreover, the gills' features could describe the variance in the storage durations more efficiently than those extracted from the eyes. Finally, it was concluded that the applied colorimetric system along with the developing models could be employed as a successful non-destructive approach for evaluation of fish freshness.