Underwater Moving Object Localisation Based on Weak Electric Fish Bionic Sensing Principle and LSTM

Traditional underwater target object detection technologies such as sonar and optical imaging can’t be well applied to harsh underwater environment. Inspired by the biological perception of weak electric fish, an underwater moving object localisation method based on active electric field sensing principle and Long and Short Term Memory (LSTM) network is proposed in this paper. A sensor array with 4 electric Held emitting electrodes and 16 electric field receiving electrodes is designed, and LSTM network with 6- layer architecture is constructed. The pool experiment results show that the predicted object motion path of the LSTM network is very close to the real path. In a 500mm×80mm area of interest, the two-dimensional mean absolute localisation error is 5.38mm and the two-dimensional mean relative localisation error is 1.06%, which proves that the proposed method has tiny localisation error and high localisation accuracy for underwater moving object.