This research was conducted to enable the in-situ measurement of the aspect and grazing angle dependent scattering response of unexploded ordnance (UXO), clutter, and science targets by conducting circular scans at different heights around identified concentric and non-concentric (i.e., off-center) targets. Methods were developed for coherently fusing the information from all circular scans to generate three dimensional (3D) data products describing spatial reflectivity and impulse response of targets. Showing that these 3D dataproducts could be sampled in the 3D wavenumber spectrum to generate a series of two dimensional (2D) random orientation realizations that have use for target recognition training and/or testing was an additional objective.
The primary signal processing challenges of this project were 1) the precision, data-driven alignment of synthetic aperture data captured over a very large multi-dimensional array consisting of multiple circles or, in the final implementation, a corkscrew like spiral aperture, and 2) the coherent fusion of the data captured by the multi-dimensional array into a three-dimensional data-product from which useful spatial and frequency response information about targets could be gathered. To address these challenges a crawl-walk-run approach was taken. In the first year highly controlled multi-pass circular scans were conducted via a sonar system mounted on a tower and traversing a circular rail in a controlled pond environment. Basic multi-pass data fusion (e.g., 3D beamforming) and concentricity correction algorithms were developed using this dataset. In the second year a dual band sonar system mounted on a REMUS 600 autonomous underwater vehicle (AUV) conducted multipass circular scans of varying types in both benign and complex uncontrolled ocean environments. Finally, in the last year of the project the same sonar system was used to test a more efficient means of conducing multipass circular scans. Using this method a much larger sampling of the multidimensional scattering response of targets could be captured in the same amount of time as in previous methods, but the signal-processing approach for fusing the data had to be completely reworked.
Results demonstrated that not only were 3D multipass synthetic aperture sonar feasible, but were extremely doable in field conditions and for objects the size of UXO. 3D impulse response data showing both rigid and elastic structural response information was readily extractable from 3D low-frequency image data products. It was further demonstrated that by appropriately sampling the 3D wavenumber spectrum, realizations of the target from random orientations could be generated.
3D multipass processing has a number of potential uses, ranging from high certainty target and environment classification and characterization to classifier training and performance analysis. Additionally, many of the 3D synthetics aperture sonar (SAS) signal processing techniques developed in this project (e.g., 3D autofocus procedures and 3D region-of-interest impulse response extraction) could potentially be extended to work with 3D downlooking SAS systems used for funding buried targets.