Underwater sound target identification method based on morphological component analysis feature fusion

Abstract

The invention discloses an underwater acoustic target recognition method based on morphological component analysis feature fusion, and the method comprises the following steps: S1, carrying out the preprocessing of the radiation noise of an underwater acoustic target, and extracting a feature image of the target; s2, for the target feature image extracted in the S1, establishing a sparse model of the target feature image; s3, performing feature component separation on the feature sparse model established in the S2, and constructing a G feature of the target by using the separated feature components; s4, utilizing the feature components separated in the step S3 to respectively train to obtain a public redundancy dictionary of sparse representation of each component; s5, designing a feature fusion framework based on morphological component analysis, and constructing fusion features of the target through the fusion framework; and S6, designing a convolution-based recognition network, and training the recognition network by using the fusion features in the S5 to obtain a high-accuracy recognition network, thereby realizing recognition and classification of targets with similar features, and improving the recognition accuracy.

Description

technical field [0001] The invention relates to the field of underwater acoustic target recognition, in particular to an underwater acoustic target recognition method based on morphological component analysis feature fusion. Background technique [0002] With the emergence of sonar technology, underwater acoustic target recognition technology, which is an important follow-up link of sonar data processing, also emerged as the times require. Underwater acoustic target recognition refers to the information processing technology that uses the radiation noise of the target to extract characteristic information and perform target discrimination and classification. At present, there are mainly two types of target recognition methods commonly used. The first type is statistical classification recognition, which mainly uses the statistical distribution of target features and relies on statistical analysis of existing sample data and pattern matching based on distance metrics; the sec...

AI Technical Summary

Problems solved by technology

However, as the sounding mechanism of underwater acoustic targets becomes more and more complex, the types of targets change and the interference of environmental noise causes the characteristic information of the target to become weaker and weaker, it is difficult to be effectively extracted, and it becomes a constraint that limits target recognition.

In recent years, nonlinear feature extraction technology has developed rapidly. At present, the main methods include Hilbert-Huang transform, wavelet analysis, high-order spectrum analysis and other modern signal processing feature extraction methods, but the related algorithms are too cumbersome and can only be used for It is difficult to fully characterize the characteristics of the target by extracting the local feature information of the target in a targeted manner

For targets with similar characteristics, it is easy to generate discrimination errors, resulting in a decrease in the accuracy of target recognition and an increase in the false alarm rate.

[0004] At the same time, in underwater acoustic target recognition, the common feature fusion is the splicing and combination of different feature components. Although the feature information content of the target is increased, the amount of feature data is also increased, and the overlapping features Some redundant information will appear, which increases the difficulty of target recognition and causes a waste of computing resources

Images