A remote sensing image change detection method based on progressive perception

By employing a progressive sensing remote sensing image change detection method, utilizing feature decoupling and defect-guided decoupling distribution incremental units, combined with a progressive control loss function, the method addresses the insufficient generalization ability of remote sensing image change detection in complex scenarios, and achieves high-precision identification of change areas.

CN122435484APending Publication Date: 2026-07-21ZHONGBEI UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
ZHONGBEI UNIV
Filing Date
2026-04-17
Publication Date
2026-07-21

AI Technical Summary

Technical Problem

Existing remote sensing image change detection methods lack generalization ability when facing complex scenes, especially when the lighting conditions change dynamically, the model performance deteriorates, traditional methods have poor accuracy, and deep learning methods have limitations in long-distance information relationship modeling and cross-domain adaptability.

Method used

A remote sensing image change detection method based on progressive perception is adopted. By using feature decoupling and defect-guided decoupled distribution increment units, multi-scale convolutional feature separators and prior distribution reparameterization sampling are used to gradually locate the change region. The model performance is optimized by combining progressive control loss function.

Benefits of technology

It achieves high-precision change detection in complex remote sensing image scenarios, improves the model's generalization ability and detection accuracy, and can effectively identify the main body and edge details in the changed area.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122435484A_ABST
    Figure CN122435484A_ABST
Patent Text Reader

Abstract

The present application belongs to the technical field of remote sensing image change detection, and particularly relates to a remote sensing image change detection method based on progressive perception. In order to solve the problem of poor detection accuracy in complex scenes, the present application: firstly, obtains a decoupling feature posterior probability distribution through a difference feature and a label, and utilizes the same to train a decoupling feature prior probability generator; secondly, realizes feature decoupling and generates change detection results of each stage by combining a multi-scale convolution feature separator and a prior distribution reparameterization sampling; then, updates the decoupling feature prior distribution generator for a change region (including missed detection and false detection) that is not correctly detected, and repeats the process in a multi-stage manner to refine the change region; finally, an incremental control loss is adopted to induce the model to gradually define the functions of each sub-module and improve the overall performance of the model.
Need to check novelty before this filing date? Find Prior Art