Autonomous learning sonar target individual identification method

A technology of self-learning and recognition methods, applied in character and pattern recognition, sound wave re-radiation, instruments, etc., can solve the problems of low data quality, insufficient model generalization ability, and few data samples, etc., and achieve fast update speed, Good individual recognition performance, the effect of occupying less computing resources

Pending Publication Date: 2022-05-06
THE 715TH RES INST OF CHINA SHIPBUILDING IND CORP
View PDF0 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a self-learning sonar target individual identification method to solve the problem of insufficient model generalization ability caused by the lack of data samples and low data quality in the sonar target individual identification based on deep learning in the background technology question

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Autonomous learning sonar target individual identification method
  • Autonomous learning sonar target individual identification method
  • Autonomous learning sonar target individual identification method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0023] Figure 1 Is the work flow chart of the embodiment of the invention. reference Figure 1 , an autonomous learning sonar target individual recognition method, which includes the following five steps.

[0024] (1) Model acquisition steps: build an individual recognition network model for sonar target, train the model with historical sonar target database, and obtain the initial model.

[0025](2) Construction steps of template library: use the initial model to extract the corresponding historical individual feature template from the historical sonar target data, and construct the template library.

[0026] (3) Template matching step: for each current sonar data obtained, use the initial model to extract the current individual feature template corresponding to the current sonar data, and search and match in the template library to obtain the target individual recognition result.

[0027] (4) In the data updating step, the current individual feature template is stored in the tem...

Embodiment 2

[0052] (1) Acquiring a historical sonar target database, the historical sonar target database includes N historical sonar target data, each historical sonar target data includes historical sonar data and corresponding individual labels, and preprocessing each historical sonar data to generate a target spectrum (LOFAR spectrum, demon spectrum or power spectrum); The individual recognition network model based on the perception RESNET model is constructed. The historical target spectrum and individual label corresponding to the historical sonar data are used as the target input and target output of the individual recognition network model respectively. The individual recognition network model is trained. After the individual recognition model converges, the trained model y is obtained and used as the initial model. It should be noted here that although the output of the individual identification network model is the individual label, the feature template extracted in the subsequent o...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses an autonomous learning sonar target individual recognition method, which comprises the following steps of: constructing a deep learning network model for sonar target individual recognition, training the model by using historical data to obtain an initial model and a template library, then receiving new data, predicting the new data by using the initial model, extracting an individual feature template, and finally extracting an individual feature template; the method comprises the following steps: performing retrieval matching in a template library, realizing rapid individual identification or identity determination of a new target, updating the template library, selecting a model autonomous learning updating method according to sonar platform characteristics, selecting an updating method based on total historical data joint learning if the platform is a platform with sufficient hardware resources, and selecting an updating method based on total historical data joint learning if the platform is a platform with limited hardware. And if not, selecting an updating method based on field data incremental learning, and continuously learning new data to improve the model performance and better meet actual combat environment requirements.

Description

technical field [0001] The invention belongs to the field of sonar signal processing, in particular to a sonar signal processing method, more specifically to an autonomous learning sonar target individual recognition method. Background technology [0002] Under the background of many interferences and mutual interferences on the same platform, how to accurately identify individual targets is of great significance. Traditional sonar target individual recognition realizes individual target matching and recognition by extracting individual difference features such as line spectrum and constructing feature template. However, under the condition of low signal-to-noise ratio or multiple interference, the traditional sonar target individual recognition methods will cause the line spectrum characteristics to be polluted, the line spectrum is weak or invisible, which will lead to the serious decline of individual recognition performance. [0003] Deep learning is an important means to bre...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/62G06K9/00G06F16/23G01S15/02
CPCG01S15/02G06F16/23G06F2218/08G06F2218/12G06F18/22G06F18/214
Inventor 王青翠王方勇杜栓平罗兆瑞陈越超
Owner THE 715TH RES INST OF CHINA SHIPBUILDING IND CORP
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap