Direction Identification of Underwater Moving Target with Active Electrosense and CNN

Underwater active electrosense is bio-inspired by weakly electric fish that discharge and sense electrical information. Due to this biological mechanism, the fish’s ability of navigation and evasion are enhanced even in murky water. Previous researches about underwater active electrosense mostly focus on the localization and shape recognition of still object, but few on the detection of object in motion. This paper aims at providing a method of the direction identification of underwater moving target by active electrosense. We build a detection model to explain the principle of direction identification and design active electrosense array with corresponding system. Since the analytical method is difficult to implement, Convolutional Neural Network (CNN) in deep learning is used for identifying, and the experiment result shows that the accuracy reaches over 84%. Besides, comparative experiment is conducted to figure out the influence of sensors quantity on identification performance. This technology can be applied to underwater search and exploration in obscure environment.