We consider the challenge of tracking and estimating the size of a single submerged target in a high reverberant underwater environment using a single active acoustic transceiver. This problem is common for a multitude of applications, ranging from the security and safety needs of tracking submerged vehicles and scuba divers, to environmental research and management implications such as the monitoring of pelagic fauna. Considering that the target can be either slow (e.g., a scuba diver) or fastmoving (e.g., a shark), we avoid continuous signalling, and rely on the emission of wideband pulses whose reflection pattern are evaluated and reshaped in a time-distance matrix. As opposed to common approaches that track targets through template matching or by using tracking filters, we avoid making difficult assumptions about the target's reflection patterns or motion type, and instead perform probabilistic tracking using a constraint Viterbi algorithm, whereby detection is determined based on maximum likelihood criterion. In this process, we use the expectation-maximization (EM) approach to manage stationary reflections through distribution analysis, which otherwise may be misidentified as targets. Based on the tracked path, we then evaluate the target's size. To test our approach, we performed extensive simulations as well as eight sea experiments in different environmental settings to track both a scuba diver and a sandbar shark (Carcharhinus plumbeus). The simulation results show a tracking performance that is close to the Cramér-Rao lower bound, and the experiment results show a good trade-off between detection rate and false alarm rate for a low signal-to-clutter ratio of 5 [dB], and average tracking error of 1.5 [m] and 6.5 [m] in the detections of a scuba diver and sandbar shark, respectively. For reproducibility, we share our sea experiment data.

No abstract

The accurate detection and quantification of submerged targets has been recognized as a key challenge in marine exploration, one that traditional census approaches cannot handle efficiently. Here we present a deep learning approach to detect the pattern of a moving fish from the reflections of an active acoustic emitter. To allow for real-time detection, we use a convolutional neural network, which provides the simultaneous labeling of a large buffer of signal samples. This allows to capture the structure of the reflecting signal from the moving target and to separate it from clutter reflections. We evaluate system performance both on synthetic (simulated) data, as well as on real data recorded over 50 sea experiments in a variety of sea conditions. When tested on real signals, the network trained on simulated patterns showed non-trivial detection capabilities, suggesting that transfer learning can be a viable approach in these scenarios, where tagged data is often lacking. However, training the network directly on the real reflections with data augmentation techniques allowed to reach a more favorable precision-recall trade-off, approaching an ideal detection bound. We also evaluate an alternative model based on recurrent neural networks which, despite exhibiting slightly inferior performance, could be applied in scenarios requiring on-line processing of the reflection sequence.

We address the challenge of detecting an arbitraryshaped underwater acoustic signal. Instead of setting a detection threshold, which due to noise transients may result in a high false alarm rate, our method classifies each measured sample as either 'noise' or 'signal'. Utilizing a-priori knowledge of only the minimal duration of the signal, the decision is made using loopy belief propagation over a factor graph. Numerical simulations and sea experimental results show that our scheme achieves a favorable trade-off between the recall and false alarm rates, and noise robustness which far exceeds that of benchmark schemes.

No abstract