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

SAR target image generation method with controllable azimuth angle

A target image and azimuth technology, applied in neural learning methods, biological neural network models, instruments, etc., can solve problems such as high dimensionality, inaccurate SAR image azimuth, and CGAN's inability to converge Nash equilibrium state, and achieve accurate azimuth. Effect

Active Publication Date: 2019-10-22
UNIV OF ELECTRONIC SCI & TECH OF CHINA
View PDF5 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, when these two codes are controlled by CGAN to generate SAR images, CGAN cannot converge to the theoretical Nash equilibrium state. At the same time, the azimuth angle of the SAR image generated by the CGAN model trained based on the lab enconding feature coding method is inaccurate. Based on one-hot The dimension of the azimuth label of encoding is 360, which is relatively high, and noise with higher dimension is required

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 target image generation method with controllable azimuth angle
  • SAR target image generation method with controllable azimuth angle
  • SAR target image generation method with controllable azimuth angle

Examples

Experimental program
Comparison scheme
Effect test

example

[0126] In this example, image 3 and Figure 4 Generate the training result of the confrontational network for the condition in the present invention, wherein image 3 The change of the loss value of the discriminative model and the generative model during the training process of the CGAN model encoded for N-Progressive, Figure 4 SAR image generated for the N-Progressive coding based CGAN model. At the same time, the experimental results of CGAN based on two encoding methods, label encoding and one-hot encoding, are compared as follows: Figure 5 , Figure 6 , Figure 7 and Figure 8 shown.

[0127] According to the loss functions of the generative model and the discriminant model, after the model converges, the theoretical values ​​of the generative model and the discriminant model are respectively:

[0128] G_loss=-log[D(fake)]=-log(0.5)=0.693

[0129] D_loss=-[log(D(real))+log(1-D(fake))]

[0130] =-[log(0.5)+log(1-0.5)]

[0131] =1.386

[0132] From Figure 5 ...

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 target image generation method with a controllable azimuth angle. An azimuth angle of an SAR image is coded into a label by adopting an N-step progressive coding mode. The SAR image and the coded label are sent to a CGAN for training. After the model is stable, the CGAN generated model can generate the SAR image with a specified azimuth angle. According to the method, in the process of generating the SAR target image with the controllable azimuth angle, an azimuth angle discriminator is not needed for screening. The progressive coding can solve the problem of model collapse in the CGAN training process, and the azimuth angle of an SAR target image generated by using N-step progressive coding is more accurate. After the CGAN model converges, a large number ofSAR target images can be quickly generated. The SAR image features do not need to be extracted.

Description

technical field [0001] The invention belongs to the technical field of SAR target image generation, and in particular relates to a method for generating an AER target image with controllable azimuth angle. Background technique [0002] SAR has a wide range of applications in the fields of earth remote sensing, ocean research, resource exploration, disaster forecasting and military investigation. With the research on SAR deception technology and SAR image interpretation technology, we often face the problem of insufficient SAR image data sets. The cost of generating SAR images using the method of actual measurement is too high. Now the interpretation technology of SAR images often uses deep learning algorithms. The training of deep learning requires a large number of data sets as the training set of the model. Therefore, in practice, it is often necessary to artificially expand the data set. [0003] At present, there are also many researches on image generation models. For...

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): G06K9/00G06N3/04G06N3/08
CPCG06N3/08G06V20/13G06N3/045
Inventor 张伟王雷雷
Owner UNIV OF ELECTRONIC SCI & TECH OF CHINA
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