This paper presents the development and evaluation of a time-frequency processing technique for detection and classification of buried cylindrical targets from chirpbased parametric sonar data. The software is designed to discriminate between cylindrical targets —such as cables— of different diameters, which need to be identified as different from other strong reflectors or point targets. The method is evaluated on synthetic data generated with an acoustic scattering model for elastic cylinders for seven different diameters. The model generates characteristic responses of targets acquired by a parametric sonar system. The signal at the sonar receiver hydrophones is first windowed to reduce the data to the region of interest (buried target return). This return is then transformed using joint timefrequency transforms (we use the Wigner and Choi-Williams distributions) to produce a 2D image of the return. Dimensionality reduction and feature extraction are performed by singular value decomposition of this time-frequency image. Linear, quadratic, and Mahalanobis discriminant functions are then applied to the most significant singular values to produce the final classification. The study is carried out for various scenarios of free field response of targets as well as for responses from targets buried in sediment.
Kaminsky, E. and M. Barbu, “Classification of cylindrical targets buried in seafloor sediments,” in IEEE Region 5 Tech. Conf., Lafayette, AR, April 20-22, 2007, pp. 117–123.