Underwater cylinder recognition using machine learning with DFT-based feature vectors

Abstract

In detecting an object bottoming at the seabed, the target recognition is an important problem. Previous studies were based on energy detection. In this case, a false detection usually occurs when the clutter energy is larger. In this paper, we try to improve the performance of target recognition using the logistic regression model under a shallow water environment. The goal of this study is to develop a logistic regression model for recognition of a target by training the model using the DFT values of the signal received by the active sonar. The performance of the proposed method is verified through simulation experiments in terms of the target to clutter ratio.