A Smart Cane Based on 2D LiDAR and RGB-D Camera Sensor-Realizing Navigation and Obstacle Recognition

Area of Science:

• Robotics and Artificial Intelligence
• Assistive Technology
• Computer Vision

Background:

• Visually impaired individuals face significant challenges in independent navigation.
• Existing assistive technologies often lack comprehensive environmental perception and real-time guidance.
• The integration of advanced sensing and AI offers potential for enhanced mobility solutions.

Purpose of the Study:

• To develop and evaluate an intelligent blind guide system integrated into a smart cane.
• To enable autonomous navigation and obstacle avoidance for visually impaired users.
• To provide a reliable and intuitive mobility aid mimicking the function of a guide dog.

Main Methods:

• Utilized a smart cane equipped with 2D LiDAR, RGB-D camera, IMU, and GPS.
• Implemented the Cartographer algorithm for simultaneous localization and mapping (SLAM).
• Employed an improved YOLOv5 algorithm for real-time identification of various obstacles and environmental features.

Main Results:

• The system achieved a laser SLAM mapping and positioning accuracy of 1m ± 7cm with a speed of 25-31 FPS.
• The improved YOLOv5 model identified 86 object types, with recognition rates of 84.6% for crosswalks and 71.8% for vehicles.
• The intelligent guide system demonstrated effective obstacle avoidance and navigation in both indoor and outdoor environments.

Conclusions:

• The developed smart cane system provides a self-leading blind guide function, enhancing mobility for the visually impaired.
• The combination of SLAM and advanced object recognition enables efficient navigation and obstacle avoidance.
• This technology offers a promising alternative to traditional mobility aids, improving independence and safety.