Feature-Based Squids Classification Using Fuzzy Logic

Abstract

Squid is the most diversified member of marine cephalopod mollusk and it is a highly priced crustacean. The authenticity of species is a big issue for food authorities, while given wrong species labels representing the commercial fraud to the consumers, hence this implies substitutiion of one species by another of cheap commercial value. Each squid has got its own features to accurately classify the species. Color, shape, and texture features are very important in classifying the squid species. The most dramatic differences were observed in the shape variation of squid species, which are mantle length, mantle width, head length, head width, fin width, and total length. Not only the shape features are sufficient to identify the squid species but the color and textural features are also considered for classification. The color and texture features of squids are correlation, homogeneity, entropy, R mean, standard deviation, G mean, standard deviation, skewness, B mean standard deviation, and skewness of B. While considering the color, shape, and texture features of squid images, some uncertainty can be found on these features. To avoid uncertainty soft computing techniques have become more popular. In different soft computing paradigms, more flexible and uncertainty-avoiding method is fuzzy logic. Hence, in this work, we developed a feature-based fuzzy inference system design to classify squids into 15 classes. The classifier in this case operates as a Mamdani type fuzzy inference system used for the classification of squid species.