There is a need for technologies that can rapidly and accurately assess large underwater areas for detection and classification of unexploded ordnance (UXO). The objective of this SERDP Exploratory Development (SEED) project is to develop acoustic signal processing methodologies that will facilitate two tasks: (1) identifying areas with munitions concentration from areas free of munitions with high area coverage rates and low false rates, and (2) classifying targets with high probability of correct identifications. This will be done by analyzing broadband backscattered acoustic echoes collected using synthetic aperture sonar (SAS). The algorithm uses temporal features that are selected based on understanding of the biological sonar of dolphins.
This project will develop signal processing software to be used with SAS for semi-real-time UXO assessment. It will leverage work from SERDP project MR-1665 led by the Applied Physics Laboratory (APL), University of Washington, which studied the characteristics of eleven munitions targets in a pond measurement. It has been shown that the high-energy regions in the range/cross-range graphs generated by SAS implied the detection of munitions targets. This project will develop an automatic algorithm to detect munitions based on the SAS data that will be supplied by APL so that it can be done without human interventions. This is followed by analyzing the echoes with the target in the range of sonar. Parameters that broadly characterize the echo structures will be extracted from the time domain signal to quantify it from three aspects: the energy concentration, the highlights, and the exponential curve fitting of the signal envelope. Two classification algorithms will be investigated: the fuzzy c-means clustering (FCM) algorithm and the multi-class support vector machine (MC-SVM) algorithm. Data collected from SERDP project MR-1665 will be used to build the acoustic features library and to test the performance of the classification algorithm.
The system would benefit many Department of Defense sites for UXO assessment. The automated, semi-real-time detection/classification provides the capability to process the data in the field. It will significantly improve the usability of the sonar system for underwater munitions management. This research will also enhance the scientific understanding of dolphin sonar, by developing and testing signal-processing methods imitating the biological process. (Anticipated Project Completion - 2015)