Classification of fish types using scale-invariant feature transform, bag of features and support vector machine

Abstract

There are about 32.000, the total number of living fish species. It is very diverse because 70% of earth is covered in water. Due to the large number of species, it is causing some problems to classify fish. To classify fish, it needs to observe using direct eyes, and compare them with the reference books. An approach of machine learning is needed to classify fish effectively. In this study, we use 9000 fish images of 9 species of data. Scale-Invariant feature transform (SIFT) is used as a feature extraction. Then, features need to be reduce using Bag of features (BoF). Support Vector Machine (SVM) is used as a classification. Based on the results, 200 cluster of BoF obtain 98.45% of accuracy. This result show that the combination of SIFT, BoF and SVM can classify fish well.