Buried object detection by auto-regressive pre-whitening

Abstract

An advanced signal processing technique devoted to the detection of buried objects by exploiting an active sonar system is proposed. The technique is based on the modeling of the reverberation phenomenon as an auto-regressive process. The detector consists of an adaptive pre-whitening filter and a bank of matched filters. The auto-regressive parameters are computed by a higher order statistics algorithm that works on short successive reverberation segments. No echoes of the buried target have been used to arrange the matched filters, but only echoes of the target in free water. The proposed technique has been tested with an experimental data set related to a steel cylinder deeply buried in the sea bottom, obtaining impressive results in spite of the very low signal to reverberation ratio. This work was performed owing to the European Commission support, in the context of the contract DEO (Detection of Embedded Objects).