Underwater Acoustic Sensing for Target Detection with Fractional Scattering Network and Wavelet Neural Network.

Underwater target sensing is an essential topic for many civilian and military applications. Vision-based sensing techniques have limitations due to the high attenuation of electromagnetic waves in water. In this paper, acoustic sensing is applied for underwater target detection. A new neural network-based algorithm is developed for feature extraction and the classification of underwater target mobilities including the motion speed and moving directions. The fractional scattering neural network, FrScatNet is applied for feature extraction, followed by Wavelet Neural Network (WNN) for classification. To facilitate sufficient training data, a Geometric Ray Tracing Model is adopted for dataset generation. Different target sensing scenarios are simulated with varied target velocities and directions. The effectiveness of the proposed algorithm is validated by sets of simulation experiments and the evaluation of overall system performance