K-distribution reverberation and reliable statistical approach study of sonar data

Performing reliable target detection in reverberation background is a critical problem demanding prompt solutions. Researching on the statistical characterization of reverberation envelope is significantly important for not only the CFAR detection, but also the environment detection and seafloor classification. Identified as a robust statistical model for shallow water reverberation envelope, K-distribution, which is introduced from radar clutter processing to sonar engineering, provides a physical interpretation of its parameters which are relevant to the environment parameters. Because different sonar data processing methods may lead to different results, we have to research on the reliable statistical approaches of processing sonar data, in order to obtain objective and credible conclusions. In this paper, we generate pseudo-random variables of K-distribution and introduce a series of statistical procedures to handle the random data. Then the distribution fitting, the parameter estimation of the distribution and the evaluation of fitting goodness could be made on the basis of these steps. Supported by an underwater experiment, we sampled a set of real data by a planar array. We illustrate the approach using the real data, and verify that the reverberation envelope of shallow water fits the K-distribution model well.