Research News: Poisson multi-Bernoulli mixture filtering with an active sonar using BELLHOP simulation
Alexey Narykov, post doc at Signal Processing Group, provides an overview of his research paper 'Poisson multi-Bernoulli mixture filtering with an active sonar using BELLHOP simulation'
Summary
My work is concerned with processing underwater sonar detections using advanced Bayesian algorithms that might have been overlooked or deemed inapplicable by the sonar community. To alleviate the challenge of modelling the sonar operation, we employ the BELLHOP model, which is an industry standard for simulating the propagation of acoustic signals in the underwater environment. In the current publication, we model sonar detection performance using BELLHOP and communicate this information to a Bayesian multi-target tracker to enhance the processing of sonar detections.
Importance of the research
Sonar’s ability to detect targets can vary greatly depending on a target’s location within the area of surveillance. This produces regions of compromised coverage where targets stop being detectable and thus may escape operator’s awareness while still residing in the area of regard. Advanced Bayesian algorithms can take such state-dependent detection performance into account and reason that a target may have entered the region of compromised coverage, and not simply vanished from the surveillance area once detections stopped arriving. However, for this the algorithm needs to be provided with an appropriate model of detection. Unfortunately, the factors affecting the detection performance are not limited to the configuration of sonar hardware, but also include the underwater environment in which sonar signals propagate. Evaluating the environmental effects commonly requires extensive sonar domain knowledge and involved mathematical modelling, but in this work is circumvented by employing the BELLHOP model. When compared to (incorrectly) modelling the detection performance as uniform, using realistic detection performance enhances resulting tracking performance: it allows for timely initiation of tracks when false alarms are present, and it improves track continuity over the regions of compromised coverage; and ultimately promotes the utility of sonar hardware.
What comes next?
Sonar system informs the sensor operator about the presence of targets in the surveillance area, and measures quantities that help localise targets within this area and estimate other parameters of interest, such as target velocities. The underwater environment adversely affects both the detection process, as well as the quality of quantities being measured. For example, measuring the sonar signal’s return travel time provides information about the target’s range. However, when signals do not travel in straight lines, this results into a significant bias in range calculations. Addressing issues of this kind, by considering the physics of signal propagation, is the next step for our research.
This article belongs to the CDT's Fusion 2022 series. Please review our other Fusion conference paper overviews.