Classification of Sonar Targets using Support Vector Machine

摘要

The support vector machine (SVM) is a typical binary classifier having the global optimal solution because it deals with a quadratic optimization problem in the training process. In this paper we apply the SVM with a radial basis function to the classification of sonar targets and evaluate the recognition performance depending on the parameters of the SVM. The active sonar data set was obtained from the UCI machine learning repository. This data set consists of 208 patterns of active sonar returns, 111 of metal cylinder returns and 97 of rock returns having similar shapes. To find the optimum parameters of SVM empirically, we use a grid-search method and performance is evaluated in aspect-angle dependent experiment and aspect-angle independent experiment.