Clutter Discrimination Techniques for High-Resolution Active Sonar

Traditional active acoustic processing in a cluttered environment seeks to suppress or remove clutter in order to increase the signal-to-noise ratio of the signal passed to the operator. Typically, clutter removal is accomplished either by averaging and thresholding to remove low-level peaks in a displayed signal or image or by filtering the signal to suppress dominant clutter frequencies or remove predictable clutter components. To the extent that targets are well correlated from ping to ping while the clutter is not, or the targets of interest and the clutter occupy disjoint regions of the feature space of interest, these techniques effectively improve detection performance and reduce the workload of the sonar operator, who must attempt to identify all putative target detections. Unfortunately, in many heavily cluttered environments, only a relatively small percentage of the total clutter may be removed or suppressed without also removing or suppressing signals of interest in the data. In this case, it is necessary for the operator to successfully identify targets of interest that are effectively buried in bottom clutter, or to discriminate between a relatively small number of targets of interest and a possibly large number of false targets. The specific objective of this research is to develop discrimination techniques and methodologies to distinguish between targets and clutter in high-resolution acoustic sonar systems by attempting to characterize the clutter and classify objects on or near the bottom as either natural (or environmental) or manmade. Our approach to the clutter discrimination problem will be to apply several novel pattern recognition techniques based on results from the fields of multiresolution analysis and fractal image compression.