Presentation of a Classification Method for Sonar Marine Targets Using Multi-Class Support Vector Machine Algorithms

Using sonar is a suitable method for detecting and tracking underwater targets due to take advantage of the audio signals. On the other hand, precise and accurate classification of the targets is essential for making correct decision. The method which presented in this paper for classification of the marine targets is multi-class support vector machine algorithms which separate the targets. Classification features such as target strength and object speed are used for desired targets such as diver, fish, torpedoes and submarines. In this paper, based on these features and using multi-class algorithms such as: one vs. one, one vs. all and direct acyclic graph, desired targets are separated from each other and these results have been simulated using MATLAB. Finally, the performance of the mentioned algorithms has been evaluated, based on two criteria of classification accuracy and time.