Scholarlink

<p>The integration of diverse technological approaches in the unmanned aerial vehicle (UAV) navigation systems has demonstrated significant improvements in operational efficiency and autonomous flight capabilities, significantly reducing dependence on ground-based control infrastructure. This study proposes a lightweight UAV navigation algorithm that synergistically combines visual-reference coupling with an enhanced Extended Kalman Filter-based Simultaneous Localization and Mapping (EKF-SLAM) framework, addressing two critical challenges in resource-constrained scenarios: the excessive computational overhead of conventional visual SLAM implementations and the persistent issue of localization drift in feature-deprived environments. The key innovations include: (1) a dynamic visual-reference coupling mechanism that improves feature matching robustness while reducing environmental dependency by 38% compared to conventional approaches; (2) an enhanced EKF-SLAM architecture utilizing parallel computation and sparse matrix optimization techniques, achieving a 42% acceleration in feature point processing; (3) a resource-efficient neural network denoising module with sub-1MB memory footprint, specifically designed for real-time operation on low-power embedded platforms. Experimental validation demonstrates a localization accuracy of 0.15 m in feature-scarce environments, accompanied by 15.2% improvement in visual processing precision and 7.2–14.8% enhancement in data coupling capability. These technological advancements collectively confirm the system's practical applicability for embedded UAV implementations while establishing an extensible framework adaptable to various mobile robotic applications requiring computationally efficient, high-precision navigation solutions.</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

SLAM navigation algorithm for lightweight UAV based on vision-reference coupling

  • 期刊论文

  • 发表日期：

    2025年9月16日

查看全文
  • Jing Zhang,
  • Xiaofei Wang

摘要

The integration of diverse technological approaches in the unmanned aerial vehicle (UAV) navigation systems has demonstrated significant improvements in operational efficiency and autonomous flight capabilities, significantly reducing dependence on ground-based control infrastructure. This study proposes a lightweight UAV navigation algorithm that synergistically combines visual-reference coupling with an enhanced Extended Kalman Filter-based Simultaneous Localization and Mapping (EKF-SLAM) framework, addressing two critical challenges in resource-constrained scenarios: the excessive computational overhead of conventional visual SLAM implementations and the persistent issue of localization drift in feature-deprived environments. The key innovations include: (1) a dynamic visual-reference coupling mechanism that improves feature matching robustness while reducing environmental dependency by 38% compared to conventional approaches; (2) an enhanced EKF-SLAM architecture utilizing parallel computation and sparse matrix optimization techniques, achieving a 42% acceleration in feature point processing; (3) a resource-efficient neural network denoising module with sub-1MB memory footprint, specifically designed for real-time operation on low-power embedded platforms. Experimental validation demonstrates a localization accuracy of 0.15 m in feature-scarce environments, accompanied by 15.2% improvement in visual processing precision and 7.2–14.8% enhancement in data coupling capability. These technological advancements collectively confirm the system's practical applicability for embedded UAV implementations while establishing an extensible framework adaptable to various mobile robotic applications requiring computationally efficient, high-precision navigation solutions.