多传感器融合系统助视障者在复杂户外环境自主行走
Multimodal Perception System for Real Open Environment
- 融合摄像头、超声波、GPS等多模态传感器与视觉算法
- 在复杂场景中实现路径规划与障碍物规避，支持独立行走
- 适合视障辅助设备研发与智能导航系统应用
本文提出一种面向真实开放环境的新型多模态感知系统。该系统集成嵌入式计算平台、摄像头、超声波传感器、GPS及惯性测量单元（IMU）。与传统框架不同，本系统通过融合多传感器数据与先进计算机视觉算法，提升用户在户外环境中的安全行走能力。系统可高效完成导航至指定位置、穿越障碍区域及通过交叉路口等任务。特别地，采用超声波传感器与深度相机增强障碍物检测性能。路径规划模块基于实时反馈与用户当前状态，生成局部最优路径。为评估系统性能，我们在多种场景下设计实验。结果表明，该系统可在复杂环境中有效辅助用户独立、高效行走。
原文摘要 · Abstract (English)
This paper presents a novel multimodal perception system for a real open environment. The proposed system includes an embedded computation platform, cameras, ultrasonic sensors, GPS, and IMU devices. Unlike the traditional frameworks, our system integrates multiple sensors with advanced computer vision algorithms to help users walk outside reliably. The system can efficiently complete various tasks, including navigating to specific locations, passing through obstacle regions, and crossing intersections. Specifically, we also use ultrasonic sensors and depth cameras to enhance obstacle avoidance performance. The path planning module is designed to find the locally optimal route based on various feedback and the user's current state. To evaluate the performance of the proposed system, we design several experiments under different scenarios. The results show that the system can help users walk efficiently and independently in complex situations.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。