Accès libre

Conditional GAN-based Remote Sensing Target Image Generation Method

Haoyang Liu

Haoyang Liu

School of Computer Science and Engineering Xi’an Technological UniversityXi’an, China
Recherchez cet auteur sur
Sciendo|Google Scholar
Liu, Haoyang
, 
Zhiyi Hu

Zhiyi Hu

Engineering Design Institute Army Research LoboratoryBeijing, China
Recherchez cet auteur sur
Sciendo|Google Scholar
Hu, Zhiyi
, 
Jun Yu

Jun Yu

School of Computer Science and Engineering Xi’an Technological UniversityXi’an, China
Recherchez cet auteur sur
Sciendo|Google Scholar
Yu, Jun
 et 
Shouyi Gao

Shouyi Gao

School of Computer Science and Engineering Xi’an Technological UniversityXi’an, China
Recherchez cet auteur sur
Sciendo|Google Scholar
Gao, Shouyi
  
11 janv. 2021
Conditional GAN-based Remote Sensing Target Image Generation Method's Cover Image
À propos de cet article
Article précédent
Article suivant
Citez
Télécharger la couverture
Publié en ligne: 11 janv. 2021
Pages: 66 - 74
DOI: https://doi.org/10.21307/ijanmc-2020-039
Mots clés
© 2020 Haoyang Liu et al., published by Sciendo
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Figure 1.

Conditional generative countermeasure network structure
Conditional generative countermeasure network structure

Figure 2.

Working principle of marine background condition generated countermeasure network model
Working principle of marine background condition generated countermeasure network model

Figure 3.

U-net generator network structure and coding and decoding structure diagram
U-net generator network structure and coding and decoding structure diagram

Figure 4.

Network structure diagram of discriminator
Network structure diagram of discriminator

Figure 5.

Experimental process
Experimental process

Figure 6.

Preprocessing of high resolution ship remote sensing image
Preprocessing of high resolution ship remote sensing image

Figure 7.

Model training process
Model training process

Figure 8.

Sample random sample conditional mask
Sample random sample conditional mask

Figure 9.

Generator structure of contrast experiment model
Generator structure of contrast experiment model

Figure 10.

Residual block network structure in converter
Residual block network structure in converter

Figure 11.

Random conditional sample generation results
Random conditional sample generation results

COMPARISON OF PSNR EVALUATION RESULTS

Image generation networkSSIM
The method of this paper88.47%
A generation model with nine residual blocks81.65%
A generating model with six residual blocks76.31%

DISCRIMINATOR MODEL NETWORK STRUCTURE PARAMETER TABLE

InputsTypeKernelBatch NormalizationActivation FunctionOutputs
256x256conv4x4YESLeakyReLU128x128
128x128conv4x4YESLeakyReLU64x64
64x64conv4x4YESLeakyReLU32x32
32x32conv4x4YESLeakyReLU31x31
31x31conv4x4YESLeakyReLU30x30

EXPERIMENTAL ENVIRONMENT

Operating systemUbuntu 18.04 LTS 64bit
CPUIntel(R )Xeon(R) Gold 5118 CPU@2.30GHz
GPUNvidia GeForece TITAN Xp
Memory32G
programing languagePython3.6.1
compilerPycharm2018.3
Deep learning frameworkpytorch 0.4

GENERATOR MODEL NETWORK STRUCTURE PARAMETER TABLE

InputsTypeKernelBatch NormalizationActivation FunctionOutputs
256x256conv4x4YESRELU128x128
128x128conv4x4YESRELU64x64
64x64conv4x4YESRELU32x32
32x32conv4x4YESRELU16x16
16x16conv4x4YESRELU8x8
8x8conv4x4YESRELU4x4
4x4conv4x4YESRELU2x2
2x2conv4x4YESRELU1x1
1x1deconv4x4YESRELU2x2
2x2deconv4x4YESRELU4x4
4x4deconv4x4YESRELU8x8
8x8deconv4x4YESRELU16x16
16x16deconv4x4YESRELU32x32
32x32deconv4x4YESRELU64x64
64x64deconv4x4YESRELU128x128
128x128deconv4x4YESRELU256x256
Langue:
Anglais
Périodicité:
4 fois par an
Sujets de la revue:
Informatique, Informatique, autres
RSS Feed de la revue