Scholarlink

Visual SLAM (Simultaneous Localization and Mapping) integrates positioning and map building by analyzing camera-captured images, enabling precise localization and mapping for robots or drones in unknown environments. Traditional SLAM algorithms struggle with dynamic objects, reducing localization accuracy. Advances in deep learning have introduced algorithms for dynamic feature recognition, enhancing the robustness of semantic SLAM in tracking, mapping, and loop closure detection in dynamic environments. The article will cover classic SLAM methods and semantic SLAM using techniques like object detection, semantic segmentation, and instance segmentation, concluding with a summary and future outlook.

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

A Review of Visual SLAM Methods for Semantic Information Extraction Based on Deep Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Qiang Fu,
  • Fanzhi Zeng,
  • Yuanfa Ji,
  • Suqing Yan

摘要

Visual SLAM (Simultaneous Localization and Mapping) integrates positioning and map building by analyzing camera-captured images, enabling precise localization and mapping for robots or drones in unknown environments. Traditional SLAM algorithms struggle with dynamic objects, reducing localization accuracy. Advances in deep learning have introduced algorithms for dynamic feature recognition, enhancing the robustness of semantic SLAM in tracking, mapping, and loop closure detection in dynamic environments. The article will cover classic SLAM methods and semantic SLAM using techniques like object detection, semantic segmentation, and instance segmentation, concluding with a summary and future outlook.