Underwater acoustic target forward scattering acoustic disturbance positioning method based on transfer learning

Abstract

The invention relates to an underwater acoustic target forward scattering acoustic disturbance positioning method based on transfer learning. Acoustic disturbance characteristic information of forwardscattering of an underwater acoustic target is extracted, a convolutional neural network is trained and transfer learning is performed in combination with simulation data and a small amount of actualdata, a shallow sea underwater acoustic target positioning prediction model is established, and positioning of the underwater acoustic target can be realized; compared with an existing positioning method, the method can reduce the influence of environmental mismatch on the positioning performance.

Description

technical field [0001] The invention relates to a method for locating an underwater acoustic target, in particular to a method for locating forward scattering acoustic disturbance of an underwater acoustic target based on transfer learning, and belongs to the field of signal processing. Background technique [0002] With the development of underwater target noise reduction technology and noise reduction technology, target radiation noise and target intensity are gradually reduced, and it is becoming more and more difficult for traditional passive sonar and active sonar with transceivers to detect and locate underwater acoustic targets. In recent years, dual / multistatic sonar has attracted more and more attention because of its unique transceiver structure. What arrives at the receiving end is the forward and side scattered waves of the target in different directions. However, due to the aliasing of direct arrival waves and target forward scattered waves, bistatic / multistatic...

AI Technical Summary

Problems solved by technology

However, due to the aliasing of the direct arrival wave and the forward scattered wave of the target, the bistatic / multistatic sonar has a positioning blind zone near the transmitting and receiving line, and the traditional positioning method based on the direction of arrival estimation is difficult to locate the target in the blind zone

[0003] Aiming at the problem of positioning blind spots, a disturbance sound ray method is proposed, and the intersection point of multiple disturbance sound rays is used as the target positioning result, but this method is difficult to apply in the shallow sea and long-distance environment; the sensitive nuclear positioning method transforms the target positioning problem into a linear inversion problem, but the method is driven by the model and is easily affected by the environment mismatch; the positioning method using the neural network faces the problem of difficult feature information selection, and the simulation training data has a certain environment mismatch with the actual data, and the method is not robust

Images