Scholarlink

<p>To address the challenges of nonlinear radiometric distortions (NRDs) and geometric deformations in multimodal remote sensing image (MRSI) matching, this study proposes a&#xa0;novel approach that integrates shape and curvature descriptors to achieve robust correspondence establishment. First, a&#xa0;Feature Essence Extraction Strategy (EC) is developed using the Shape Factor Diagram (SFD) and the Normalized Shape Factor Diagram (NSFD). By applying adaptive thresholding, this strategy optimizes feature point detection and increases feature point repeatability by a&#xa0;factor of 1.2–3.1 compared with conventional methods. Second, curvature diagram (CD)-derived descriptors are employed om place of traditional gradient- and phase congruency-based descriptors, enhancing resilience against NRDs. Experiments conducted on 100 MRSI pairs with significant geometric and radiometric variations show that the proposed method achieves a&#xa0;matching correctness rate of 26.71% and reduces the average RMSE to 1.91 pixels. These results demonstrate the method’s superiority over existing state-of-the-art techniques for MRSI matching.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Improved Method for Multimodal Remote Sensing Image Matching Based On Curvature and Shape Factor Diagram

  • 期刊论文

  • 发表日期：

    2025年12月15日

查看全文
  • Wu Zhang,
  • Chang Wang

摘要

To address the challenges of nonlinear radiometric distortions (NRDs) and geometric deformations in multimodal remote sensing image (MRSI) matching, this study proposes a novel approach that integrates shape and curvature descriptors to achieve robust correspondence establishment. First, a Feature Essence Extraction Strategy (EC) is developed using the Shape Factor Diagram (SFD) and the Normalized Shape Factor Diagram (NSFD). By applying adaptive thresholding, this strategy optimizes feature point detection and increases feature point repeatability by a factor of 1.2–3.1 compared with conventional methods. Second, curvature diagram (CD)-derived descriptors are employed om place of traditional gradient- and phase congruency-based descriptors, enhancing resilience against NRDs. Experiments conducted on 100 MRSI pairs with significant geometric and radiometric variations show that the proposed method achieves a matching correctness rate of 26.71% and reduces the average RMSE to 1.91 pixels. These results demonstrate the method’s superiority over existing state-of-the-art techniques for MRSI matching.