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

SAR ship target detection method based on network pruning and knowledge distillation

A target detection and network technology, which is applied in the application field of radar remote sensing images, can solve the problems of large model calculation and capacity, many network parameters, and decreased detection speed.

Active Publication Date: 2021-02-02
NAT UNIV OF DEFENSE TECH
View PDF9 Cites 16 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Although the detection accuracy has been improved, the calculation amount and capacity of the model are relatively large and there are many network parameters, so the detection speed has decreased

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
  • SAR ship target detection method based on network pruning and knowledge distillation
  • SAR ship target detection method based on network pruning and knowledge distillation
  • SAR ship target detection method based on network pruning and knowledge distillation

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0050] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0051] The SAR ship target detection method based on network pruning and knowledge distillation includes the following steps:

[0052] Step 1: Input the SAR image slice into the detection network, and use the YOLOv3 detector as the reference detection framework; introduce the asymmetric convolution module ACM in the last three stages of the backbone network to enhance the feature representation ability of objects with large aspect ratios; By adding convolution kernels of different shapes to enrich the receptive field of the convolution kernel, the detection of targets of different shapes is improved; the input feature is expressed as F in , and F in The output feature maps from the last three stages in the backbone network; the input feature F in Access three branches, where the convolution kernel sizes corresponding to the first branch and t...

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 SAR ship target detection method based on network pruning and knowledge distillation, and belongs to the technical field of radar remote sensing image application. Accordingto the technical scheme, firstly, a backbone network structure suitable for SAR multi-scale and large-length-width-ratio targets is designed, and then channel pruning is conducted on the network to generate a compact model. In addition, a knowledge distillation strategy is applied to make up performance degradation caused by network pruning. Different from full-feature simulation in a common detection model distillation method, the method has the advantages that the mutual relation between feature maps of different levels is used as migration knowledge in distillation, and a novel attention mechanism is designed to enhance target-related features, so that the distilled features have higher representation capability. By adopting the detector constructed by the invention, the model size of2.8 M, the reasoning speed of more than 200fps and lower calculation cost can be realized, and the detection precision is also improved to a certain extent.

Description

technical field [0001] The invention belongs to the technical field of radar remote sensing image application, and relates to a convolutional neural network-based SAR (Synthetic Aperture Radar) image ship detection method, especially a target detection method combined with deep network compression. Background technique [0002] Synthetic Aperture Radar (SAR) is an active microwave imaging sensor, which uses range-oriented pulse compression technology and azimuth-oriented aperture synthesis technology to achieve higher spatial resolution, and uses airborne and spaceborne, etc. The platform realizes high-resolution imaging of a large-scale observation area. At present, the research on ship target detection based on SAR images has attracted the attention of various countries, and it is of great significance to safeguard maritime rights and interests, perform maritime rescue missions, and implement precise guidance of maritime targets. [0003] In recent years, convolutional ne...

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): G06K9/00G06K9/62G06N3/08G06N3/04
CPCG06N3/082G06V20/13G06V2201/07G06N3/045G06F18/214G06F18/253G06F18/24
Inventor 占荣辉陈诗琪王威刘盛启张军
Owner NAT UNIV OF DEFENSE TECH
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