Scholarlink

This study proposes an innovative indoor augmented reality navigation framework integrating laser panoramic 3D reconstruction and visual place recognition (VPR) to address three critical challenges in existing systems: meter-level localization errors (5–7 m in WIFI fingerprinting)‌, prohibitive infrastructure costs (>USD 20,000 per 1,000 m2 for UWB)‌, and limited environmental adaptability (300% magnetic interference in steel structures)‌. A hybrid methodology combining consumer-grade LiDAR mapping (0.02 m accuracy) with enhanced SuperGlue feature matching achieves sub-meter localization precision while reducing hardware costs by 82% compared to conventional solutions‌. Experimental validations in complex indoor scenarios (shopping malls, libraries, exhibition halls) demonstrate ‌sub-meter localization precision (5–10 cm)‌—surpassing UWB and WiFi fingerprinting by 90–95%. Leveraging ‌native WeChat WebAR deployment‌, the framework eliminates dependency on dedicated apps, enabling real-time mobile navigation with negligible latency. This study establishes a ‌cost-effective paradigm‌ for intelligent building navigation, balancing centimeter-level mapping accuracy with scalable edge-computing workflows.

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

Indoor AR Navigation System Based on Deep Vision

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Xinyuan Zhang,
  • Bin Liu

摘要

This study proposes an innovative indoor augmented reality navigation framework integrating laser panoramic 3D reconstruction and visual place recognition (VPR) to address three critical challenges in existing systems: meter-level localization errors (5–7 m in WIFI fingerprinting)‌, prohibitive infrastructure costs (>USD 20,000 per 1,000 m2 for UWB)‌, and limited environmental adaptability (300% magnetic interference in steel structures)‌. A hybrid methodology combining consumer-grade LiDAR mapping (0.02 m accuracy) with enhanced SuperGlue feature matching achieves sub-meter localization precision while reducing hardware costs by 82% compared to conventional solutions‌. Experimental validations in complex indoor scenarios (shopping malls, libraries, exhibition halls) demonstrate ‌sub-meter localization precision (5–10 cm)‌—surpassing UWB and WiFi fingerprinting by 90–95%. Leveraging ‌native WeChat WebAR deployment‌, the framework eliminates dependency on dedicated apps, enabling real-time mobile navigation with negligible latency. This study establishes a ‌cost-effective paradigm‌ for intelligent building navigation, balancing centimeter-level mapping accuracy with scalable edge-computing workflows.