Image hot fog removing method based on generative adversarial network

Through the image thermal mist removal method based on the generative adversarial network, the problem of thermal mist disturbance in high-temperature DIC measurement is solved, and high-precision image processing is achieved, which reduces experimental complexity and cost and expands the application range.

CN119228664AInactive Publication Date: 2024-12-31BEIHANG UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202411333284.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-24
Publication Date
2024-12-31
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In high-temperature digital image correlation method (DIC) measurement, thermal mist disturbance causes image distortion and measurement error. Although the existing technology has been optimized, it relies on specific spatiotemporal correlation models, lacks strict mathematical descriptions, and optimization measures increase experimental complexity and Cost is difficult to popularize in large-scale applications.

Method used

Using the image thermal mist removal method based on the generative adversarial network, by obtaining image data in room temperature and high temperature environments, using the DeblurGAN_V2 generator and multi-scale discriminator, the adversarial loss function is calculated, and the generator is trained to remove thermal mist interference, which is suitable for 100 -3000℃ high temperature environment white, blue and ultraviolet DIC systems.

Benefits of technology

有效降低热雾干扰,提高图像清晰度和对比度,提升DIC测量精度,减少数据波动和误差,降低实验复杂性和成本,扩展应用范围至更多高温材料和结构的实验研究。

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119228664A_ABST
    Figure CN119228664A_ABST
Patent Text Reader

Abstract

The invention discloses a high-temperature image hot fog removing method based on a generative adversarial network (GAN). On the basis of a generative adversarial network, a generator is composed of a feature pyramid network (FPN) and a MobileNet V2, and global and local information is reserved to the maximum extent through a multi-size relative discriminator, so that rapid and efficient image hot fog removal is realized. According to training data, a reference image without hot fog and a hot fog interference image under a high-temperature condition are acquired through a digital image correlation (DIC) image acquisition system composed of a camera, a lens and a light source, and a training set and a test set for model training are constructed. And performing preprocessing and data augmentation on the hot fog image, and inputting the hot fog image into the generative network to generate a hot fog removal image. And through confrontation training of the generator and the discriminator, the performance of the generator is optimized. Finally, the fully trained generator can achieve image hot fog removal in a high-temperature environment, the definition and contrast ratio of the image are improved, a DIC system can recognize and track feature points more accurately, and therefore the measurement precision and reliability are enhanced. The method can be used for improving the measurement precision of DIC full-field deformation measurement of aerospace materials in a high-temperature environment.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Generative adversarial network image defogging method fusing feature pyramid

    CN111738942A

  • Traffic image defogging method based on improved generative adversarial network

    CN112801902A