Target classification method of double-path coupling deep learning based on sound wave propagation equation

Abstract

The embodiment of the invention provides a target classification method for double-path coupling deep learning based on a sound wave propagation equation. A category to which the to-be-classified target belongs is determined by adopting a preset deep neural network classification model; the hidden layer has a two-way coupling structure and a cross-layer jump connection structure, the SWTNN classification model can have the double-path coupling capacity and the jumping connection capacity between layers at the same time, the problem of gradient disappearance can be solved, the training precision and the classification precision of the SWTNN classification model are improved, and then the efficiency and the accuracy of determining the category to which the to-be-classified target belongs are greatly improved. Moreover, the SWTNN classification model adopted in the embodiment of the invention is constructed based on a frequency domain first-order sound wave propagation equation and a finite difference algorithm, has clear physical and mathematical significance, is an interpretable deep neural network classification model, is another important technical breakthrough, and has very high innovativeness.

Description

technical field [0001] The present invention relates to the technical field of image classification processing, and more specifically, to a target classification method based on two-way coupled deep learning of the acoustic wave propagation equation. Background technique [0002] Object recognition is an important branch in the field of image classification processing technology. How to quickly and accurately realize image classification is a current research hotspot. In recent years, a large number of researchers have used neural networks to construct classification models for object classification. [0003] The neural network may include a convolutional neural network, a residual neural network (Residual Neural Network, ResNet), and the like. However, due to the inherent nature of the network structure of the existing convolutional neural network and residual neural network, the training accuracy and classification accuracy of the classification model constructed by apply...

AI Technical Summary

Problems solved by technology

However, due to the inherent nature of the network structure of the existing convolutional neural network and residual neural network, the training accuracy and classification accuracy of the classification model constructed by applying it cannot be guaranteed, and thus the efficiency and accuracy of target recognition cannot be guaranteed.

Images