Scholarlink

<p>With the increasing complexity of industrial manufacturing scenarios and the increasing precision required for tasks, the localization problem of industrial robots in dynamic environments has become increasingly prominent. Factors such as uneven lighting, noise interference, and mechanical flexibility lead to reduced image clarity and unstable feature point matching, compromising the accuracy and reliability of robot operations. To deal with this issue, this study designs an integrated localization model that utilizes a collaborative “enhancement-matching-optimization” mechanism to achieve comprehensive improvements from imaging to geometry. The model consists of three components: an adaptive constrained dark channel enhancement module to restore image contrast and color; a robust geometric localization module based on a scale-invariant feature transform algorithm, combining threshold adaptation and directional consistency constraints to stabilize feature matching; and a weighted optimization module to improve the accuracy of homography matrix and pose estimation. In trajectory experiments, the proposed method achieved an absolute trajectory error of 0.15 and an average relative pose error reduction of approximately 40% compared to traditional methods. Under 20% occlusion conditions, the success rate was 16.1% higher than that of graph optimization-based methods, while maintaining a frame rate of 21.7 FPS. The findings denote that the designed localization model exhibits robustness and accuracy, providing a viable solution and methodological support for high-precision localization of industrial robots in intelligent manufacturing environments.</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

An industrial robot positioning model based on the SIFT algorithm

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月17日

查看全文
  • Yang Zhang

摘要

With the increasing complexity of industrial manufacturing scenarios and the increasing precision required for tasks, the localization problem of industrial robots in dynamic environments has become increasingly prominent. Factors such as uneven lighting, noise interference, and mechanical flexibility lead to reduced image clarity and unstable feature point matching, compromising the accuracy and reliability of robot operations. To deal with this issue, this study designs an integrated localization model that utilizes a collaborative “enhancement-matching-optimization” mechanism to achieve comprehensive improvements from imaging to geometry. The model consists of three components: an adaptive constrained dark channel enhancement module to restore image contrast and color; a robust geometric localization module based on a scale-invariant feature transform algorithm, combining threshold adaptation and directional consistency constraints to stabilize feature matching; and a weighted optimization module to improve the accuracy of homography matrix and pose estimation. In trajectory experiments, the proposed method achieved an absolute trajectory error of 0.15 and an average relative pose error reduction of approximately 40% compared to traditional methods. Under 20% occlusion conditions, the success rate was 16.1% higher than that of graph optimization-based methods, while maintaining a frame rate of 21.7 FPS. The findings denote that the designed localization model exhibits robustness and accuracy, providing a viable solution and methodological support for high-precision localization of industrial robots in intelligent manufacturing environments.