The classification of spherical shells with varying thickness-to-radius ratios based on the auditory perceptive features

The detection and classification of underwater objects are an important part of the coast and harbor security. For silent targets, active sonar is always employed, and then, the information contained in the scattering echo is extracted and applied. Previous studies are aimed at targets with varied shapes, and the classification of objects with similar shapes has yet to be studied comprehensively. To address this issue, the classification of spherical shells with varying thickness-to-radius ratios (TRRs) is studied. The numerical results indicate that spherical shells can be roughly divided into three classes (extremely thin, thin, and thick) according to their scattering components. In this work, the scattering features, especially auditory perceptive features for spherical shells with different TRRs, are extracted and analyzed. Then, the features are imputed into a support vector machine classifier. The simulation results of the recognition rates under different features and signal-to-noise ratios will be present and analyzed. This study provides a novel concept for the classification of silent underwater targets.The detection and classification of underwater objects are an important part of the coast and harbor security. For silent targets, active sonar is always employed, and then, the information contained in the scattering echo is extracted and applied. Previous studies are aimed at targets with varied shapes, and the classification of objects with similar shapes has yet to be studied comprehensively. To address this issue, the classification of spherical shells with varying thickness-to-radius ratios (TRRs) is studied. The numerical results indicate that spherical shells can be roughly divided into three classes (extremely thin, thin, and thick) according to their scattering components. In this work, the scattering features, especially auditory perceptive features for spherical shells with different TRRs, are extracted and analyzed. Then, the features are imputed into a support vector machine classifier. The simulation results of the recognition rates under different features and signal-to-noise ratios will b...