Scholarlink

<p>Digital image correlation (DIC) delivers non‐contact, high‐precision, full‐field measurements with simple hardware. The integer‐pixel matching serves as the starting estimation for the subsequent sub‐pixel solver (e.g., IC-GN). A better integer‐pixel matching reduces the sub‐pixel iterations and stabilizes convergence. To further overcome the low efficiency and poor robustness of integer‐pixel matching in DIC under large‐angle rotations, a fast-matching method based on a novel local feature descriptor and floating-binary-hash filtering is proposed. Firstly, a more lightweight "<i>L</i>-shaped" descriptor is designed in which gray values are sampled from integer pixel locations on horizontal and vertical pixel arrays, mapping the 2D subset into a 1D gray value sequence. Secondly, a floating-binary‐hash filtering strategy is employed to accelerate matching. Key dimensions and their random thresholds are used to construct floating-binary hash tables. Descriptors are quickly assigned to hash buckets, which greatly reduces the search domain. Both simulation and actual experiments demonstrate that the combination of the proposed feature descriptor with floating-binary-hash based filtering strategy achieves good integer‐pixel matching, and can provide accurate initial values for the sub-pixel iteration process.</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

Fast Integer-Pixel Matching for DIC Under Large Rotation Based on a Novel Feature Descriptor and Floating-Binary-Hash Filtering

  • 期刊论文

  • 发表日期：

    2026年4月6日

查看全文
  • C. Zhang,
  • X. Shao

摘要

Digital image correlation (DIC) delivers non‐contact, high‐precision, full‐field measurements with simple hardware. The integer‐pixel matching serves as the starting estimation for the subsequent sub‐pixel solver (e.g., IC-GN). A better integer‐pixel matching reduces the sub‐pixel iterations and stabilizes convergence. To further overcome the low efficiency and poor robustness of integer‐pixel matching in DIC under large‐angle rotations, a fast-matching method based on a novel local feature descriptor and floating-binary-hash filtering is proposed. Firstly, a more lightweight "L-shaped" descriptor is designed in which gray values are sampled from integer pixel locations on horizontal and vertical pixel arrays, mapping the 2D subset into a 1D gray value sequence. Secondly, a floating-binary‐hash filtering strategy is employed to accelerate matching. Key dimensions and their random thresholds are used to construct floating-binary hash tables. Descriptors are quickly assigned to hash buckets, which greatly reduces the search domain. Both simulation and actual experiments demonstrate that the combination of the proposed feature descriptor with floating-binary-hash based filtering strategy achieves good integer‐pixel matching, and can provide accurate initial values for the sub-pixel iteration process.