Multi-temporal optical remote sensing image thick cloud removal model, method and device

By designing a thick cloud removal model for multi-time phase optical remote sensing images, using feature extraction, feature fusion and image reconstruction modules, the problem of time difference in multi-time phase remote sensing images is solved, and efficient thick cloud removal and information recovery is achieved, which improves the removal of the problem. Cloud performance.

CN119151824APending Publication Date: 2024-12-17KUNMING UNIV OF SCI & TECH
View PDF 4 Cites 3 Cited by

Patent Information

Application Number
CN202410989399.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-23
Publication Date
2024-12-17

AI Technical Summary

Technical Problem

The prior art cannot effectively overcome the impact of time differences in multi-time phase remote sensing images, resulting in poor information reconstructing thick cloud occlusion areas and insufficient information extraction capabilities.

Method used

A multi-time phase optical remote sensing image thick cloud removal model is designed, including feature extraction module, feature fusion module and image reconstruction module. Through nonlinear feature extraction, multi-scale feature fusion, self-attention convolutional modulation, and low-frequency information to high-frequency information The process realizes image reconstruction of thick cloud areas.

Benefits of technology

The ability to recover spectral information and retain semantic information under thick cloud occlusion has been improved, and the performance of cloud removal has been significantly improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119151824A_ABST
    Figure CN119151824A_ABST
Patent Text Reader

Abstract

The invention relates to a multi-temporal optical remote sensing image thick cloud removal model, method and device. The model comprises a feature extraction module, a feature fusion module and an image reconstruction module which are connected in sequence. The feature extraction module is used for sequentially performing nonlinear feature extraction, channel number adjustment and multi-scale feature extraction on the multi-temporal satellite image to obtain multi-scale features; the feature fusion module is used for sequentially performing scale compression, summation operation and self-attention-based convolution modulation operation on the multi-scale features to obtain global information; and the image reconstruction module is used for performing up-sampling and affine transformation on the global information in sequence to obtain modulation features, and performing low-frequency information-to-high-frequency information processing and high-frequency information reconstruction processing on the modulation features in sequence to obtain a cloudless image. According to the method, the spectral information recovery and semantic information retention capability under thick cloud shielding is improved, and the cloud removal performance is further improved.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Multi-scale multi-temporal remote sensing image cloud removal method

    CN114140701A

  • Remote sensing image cloud removal method and device based on multi-scale feature attention network

    CN115760641A

  • Generative adversarial network-based remote sensing SAR-optical image fusion cloud removal method

    CN116579943A

  • Multi-temporal remote sensing image thick cloud removal method based on deep learning

    CN117237213A

Cited By

  • Cloud removal method and system for remote sensing image

    CN119784638A

  • A cloud removal method and system for remote sensing images

    CN119784638B

  • Remote sensing image cloud removal method oriented to long time sequence cloud coverage

    CN120510055A