Underwater target recognition method based on multi-modal fusion

Abstract

The invention discloses an underwater target identification method based on multi-modal fusion. The method comprises the following steps: (1) data acquisition, sonar signal data acquisition and marineenvironment data acquisition; (2) data fusion and spectrogram extraction: fusing the marine environment data to obtain one-dimensional data, and extracting a Lojar spectrogram by a sonar signal; (3)model building: building a multi-modal fusion model based on a convolutional neural network, wherein a one-dimensional convolutional neural network is used for marine environment data, and a two-dimensional convolutional neural network is used for a Lotar spectrogram; and (4) target identification: inputting the acquired data into a corresponding network for training and testing to obtain an underwater target identification result. Marine environment data is introduced, multi-modal fusion is carried out in combination with a Lotar spectrogram, and more accurate and richer feature information can be provided for water phase target identification in combination with the marine environment data and the Lotar spectrogram.

Description

technical field [0001] The invention relates to the technical field of marine detection, in particular to an underwater target recognition method based on multimodal fusion. Background technique [0002] The 21st century is the century of the ocean. Human beings will fully understand, develop and utilize the ocean. The development of modern science and technology has brought humans into a period of large-scale exploitation and utilization of the ocean. However, the ocean environment is extremely complex, and the ocean world is full of unknowns and dangers. Humans cannot come and go freely in the ocean as they do on land. Therefore, we need to use various technologies to deepen our understanding of the ocean. The underwater unmanned vehicle is a small submersible that can replace human beings in the ocean. It can not only conduct underwater research, but also has extremely important strategic significance. For this reason, underwater vehicles need to have a complete underwa...

AI Technical Summary

Problems solved by technology

[0006] (1) There are complex situations such as environmental noise, different types of sound sources, and multiple interference sources in the actual ocean, which lead to serious interference and noise in the signal collected by sonar, so that underwater targets cannot be accurately judged

[0007] (2) The principle of time-frequency feature extraction is simple and easy to implement, but prior knowledge is required to set the parameters, and fixed parameters will make the recognition generalization of such features weak in complex marine environment conditions

Images