Bistatic Target Classification using Low-Cost Unmanned Marine Vehicles in Shallow Water
Erin Fischell | Woods Hole Oceanographic Institution
The objective of this research is to investigate acoustic characterization techniques for seabed targets that are scalable to multiple low-cost autonomous vehicles ﬁtted with simple hydrophone arrays and sources in response to the “detection, classiﬁcation, and remediation of military munitions underwater” statement of need. Speciﬁcally, the research will address the identiﬁcation of features of bistatic and multistatic scattering ﬁelds from seabed targets for autonomous target localization and characterization using unmanned marine vehicles (UMVs). The concept of the bistatic or multistatic approach is that multiple low-cost, low-power receiver vehicles may be used to cover an area and come up with basic target contact information (e.g. location, geometry) using a single acoustic source vehicle.
When a seabed target is insoniﬁed by an acoustic source, the time delayed echoes interfere in the frequency domain to produce a three-dimensional (3D) spatial radiation pattern. This ﬁeld changes based on the target aspect angle, target geometry, composition, size, and other factors. In this approach, an acoustic data collection payload including a hydrophone array is located on a UMV and a time-synchronized acoustic source is mounted on a separate UMV or at a ﬁxed location. As the receiver vehicle progresses through the environment, the acoustic reﬂections from seabed targets are captured in time-synchronized acoustic data collection on the receiving array. Two techniques will be addressed for using this data: bistatic imaging and multistatic radiation pattern classiﬁcation (Fig. 1). In bistatic imaging, a ﬁxed source insoniﬁes a target, and a UMV with a low-cost linear hydrophone array circles the target. The acoustic data is used to form an image using modiﬁed synthetic array processing techniques, with the goal of identifying features that indicate shape and composition. In multistatic radiation pattern characterization, a source is located on one UMV and the receiver on another. The received scattering strength of the target is recorded as source and receiver move around or past the target, and the resulting dependence of scattering strength versus source-target-receiver geometry is used for estimating target characteristics using machine learning. Algorithms will be developed using simulated scattering data, and tested in real-world scattering experiments using UMVs.
This work will result in novel experimental bistatic and multistatic scattering data sets that will improve understanding of target scattering physics, critical to the development of advanced munition detection and mine countermeasures missions for the DoD. Bistatic or multistatic target scattering-based seabed object classiﬁcation would be scalable to multiple low-cost, low-power unmanned underwater and unmanned surface vehicles with the objective of more rapid coverage of large areas compared to the current technology.