Scholarlink

<p>To systematically deal with the problem lacking plan of simultaneous localization and mapping (SLAM) under low-light environments, we propose a novel visual SLAM. This new visual SLAM can robust and real-time navigate over the low-light zone. In this SLAM, we have given a complete framework to face these challenges. At the framework, we propose a fast exposure algorithm to help our system to fix a picture what obtains in a low-light environment. We name this algorithm a grayscale dual-exposure. Next, illumination-invariant filters are proposed to alleviate the exposure algorithm cases a problem of optical flow method failure. Last, due to the line features are sparse under low light, EDLines of fuzzy control are used to address this issue. The framework is comprised of these sectors. Finally, EuRoC, EuRoC_darker50%, and EuRoC_darker70% are produced to measure our SLAM. Our SLAM system demonstrates significant performance improvements in extreme low-light conditions, achieving average root mean square error (RMSE) reductions of 46.18%, 32.63%, 64.53%, and 30.14% in absolute trajectory error (ATE) compared to mainstream visual SLAM frameworks (VINS-Fusion, VINS-Mono, PL-VINS, and EPLF-VINS, respectively) when evaluated on the challenging EuRoC_darker70% dataset, a benchmark representing severely degraded lighting conditions (70% illumination reduction from standard EuRoC measurements).</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

Navigating low-light environments real-time slam based on dual exposure and fuzzy control

  • 期刊论文

  • 发表日期：

    2025年12月19日

查看全文
  • Aihua Zhang,
  • Yuhao Li,
  • Jihong He

摘要

To systematically deal with the problem lacking plan of simultaneous localization and mapping (SLAM) under low-light environments, we propose a novel visual SLAM. This new visual SLAM can robust and real-time navigate over the low-light zone. In this SLAM, we have given a complete framework to face these challenges. At the framework, we propose a fast exposure algorithm to help our system to fix a picture what obtains in a low-light environment. We name this algorithm a grayscale dual-exposure. Next, illumination-invariant filters are proposed to alleviate the exposure algorithm cases a problem of optical flow method failure. Last, due to the line features are sparse under low light, EDLines of fuzzy control are used to address this issue. The framework is comprised of these sectors. Finally, EuRoC, EuRoC_darker50%, and EuRoC_darker70% are produced to measure our SLAM. Our SLAM system demonstrates significant performance improvements in extreme low-light conditions, achieving average root mean square error (RMSE) reductions of 46.18%, 32.63%, 64.53%, and 30.14% in absolute trajectory error (ATE) compared to mainstream visual SLAM frameworks (VINS-Fusion, VINS-Mono, PL-VINS, and EPLF-VINS, respectively) when evaluated on the challenging EuRoC_darker70% dataset, a benchmark representing severely degraded lighting conditions (70% illumination reduction from standard EuRoC measurements).