Lightweight Design Method for Multi-angle SAR Target Recognition Network

A light-weight design and target recognition technology, applied in the cross-field of computer vision and remote sensing, can solve the problems of huge parameters and calculation requirements

Active Publication Date: 2022-05-10
BEIJING UNIV OF CHEM TECH
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the success of CNNs relies on the huge number of parameters and computational requirements

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
  • Lightweight Design Method for Multi-angle SAR Target Recognition Network
  • Lightweight Design Method for Multi-angle SAR Target Recognition Network
  • Lightweight Design Method for Multi-angle SAR Target Recognition Network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0037] The implementation process and experimental results of the present invention will be further described below in conjunction with experimental examples.

[0038] The sample data used in the implementation of the present invention comes from the MSTAR database published by the Moving and Stationary Target Acquisition and Recognition (MSTAR) program. The experiments used are ten types of military vehicle targets collected in the X-band and HH polarization mode, and the imaging resolution is 0.3m. The attitude coverage range of the target is 0°~360°. We take the images taken at a pitch angle of 17° as the training set, and the images taken at a pitch angle of 15° as the test set. The settings of the data set are shown in Table 1.

[0039] Table 1. Training and testing datasets

[0040] type Training set (17°) Test set (15°) 2S1 299 274 BMP2 233 195 BRDM2 298 274 BTR70 233 196 BTR60 256 295 D7 299 274 T72 232 196 ...

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 lightweight design method for a multi-angle SAR target recognition network, which belongs to the intersecting field of computer vision and remote sensing. The existing neural network compression method usually compresses ideas step by step and has the loss of recognition accuracy, while the present invention combines the requirements of multi-angle feature preservation of SAR targets, uses structured pruning to generate lightweight SAR target recognition CNN network structure, and uses knowledge distillation Restore the multi-angle feature extraction ability of the CNN network model, use weight sharing to further compress the storage space requirements of the network model, and finally obtain a lossless lightweight multi-angle SAR target recognition network model. Under the premise of no loss of recognition accuracy, the compression rate can reach more than 60 times, and the amount of calculation can be reduced by more than 2 times.

Description

technical field [0001] The invention relates to a lightweight design method for a multi-angle SAR target recognition network, which belongs to the intersecting field of computer vision and remote sensing. Background technique [0002] Synthetic Aperture Radar (SAR) is an active imaging radar, and the data generated by it has the characteristics of all-day, all-weather, high resolution and high penetration. These excellent characteristics make it widely used in geological exploration, marine monitoring, agricultural and forestry monitoring, military reconnaissance and other fields. [0003] Due to the difference in imaging mechanism, SAR images are very different from ordinary optical images, and SAR images contain a large number of special multiplicative noises, which have caused great interference to image interpretation. The interpretation methods of SAR images are mainly divided into manual interpretation and computer pattern recognition interpretation. Manual judgment ...

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 & AuthorityPatents(China)
IPC IPC(8): G06V20/00G06V10/774G06V10/762G06V10/82G06K9/62G06N3/04
CPCG06V20/13G06N3/045G06F18/23G06F18/214
Inventor张帆刘颖冰周勇胜尹嫱洪文
OwnerBEIJING UNIV OF CHEM TECH