Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Sonar image automatic target identification method based on neural network visualization

A technology for automatic target recognition and sonar images, applied in neural learning methods, biological neural network models, neural architectures, etc., can solve problems affecting generalization ability, increase method labor costs, etc., and achieve the effect of reducing labor and time costs

Inactive Publication Date: 2021-06-29
ZHEJIANG UNIV
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Most methods complete the automatic target recognition task in two steps through different techniques, which increases the labor cost of the method and also affects the generalization ability.

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
  • Sonar image automatic target identification method based on neural network visualization
  • Sonar image automatic target identification method based on neural network visualization
  • Sonar image automatic target identification method based on neural network visualization

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0032] The present invention will be described in further detail below with reference to the accompanying drawings.

[0033] Such as figure 1 As shown, the sonar image automatic target recognition method based on neural network visualization proposed by the present invention mainly includes four steps,

[0034] 1) Use the ResNet-18 classification backbone network and the Grad-CAM neural network to visualize the automatic target recognition model; where the Grad-CAM module is placed in front of the fully connected layer of the ResNet-18 classification backbone network;

[0035] 2) Construct a shape preference dataset based on the ImageNet optical dataset and adaptive instance regularization style conversion technology, use the shape preference dataset to pre-train the automatic object recognition model, and obtain pre-training parameters that are robust to shape features,

[0036] 3) Utilize the sonar image marked with the sample category as the training set to retrain the aut...

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 a sonar image automatic target identification method based on neural network visualization. By implementing the method for positioning and identifying the sonar image target only depending on the sample category labels, the labor cost of the method and the generalization ability in the field of sonar images are greatly reduced. In order to solve the problem of obvious overfitting of a training model caused by lack of a sonar data set, an optical data set used for original pre-training is converted into an optical data set based on shape preference through an adaptive instance regularization (AdaIN) style conversion method, so that the obtained pre-training parameters are more robust to shape features. Therefore, the auxiliary model focuses on extraction of specific shape features of a sonar image target in training of a sonar data set. Experiments prove that the method not only helps to solve the problem of model positioning misalignment caused by insufficient sonar data sets, but also further improves the effect of the model in the automatic target recognition task of the sonar image.

Description

technical field [0001] The invention belongs to the field of sonar target recognition, in particular to an automatic target recognition method for sonar images based on neural network visualization. Background technique [0002] The automatic target recognition technology of sonar images can not be affected by water quality and optical visibility, and is widely used in AUVs to undertake some measurement, detection and detection tasks. The automatic recognition of targets in sonar images is usually divided into two steps: positioning and recognition. The purpose of the localization part is to locate the area most likely to contain the target, while the classification part uses the information of the localization area to determine the category of the target. Most methods complete the automatic object recognition task in two steps through different techniques, which increases the labor cost of the method and also affects the generalization ability. [0003] In recent years, a...

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
IPC IPC(8): G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06N3/045G06F18/24
Inventor 郑荣濠楼冠廷
Owner ZHEJIANG UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com