AI-Powered Walking Assistance System for the Visually Impaired: Estimating Navigable Areas Using Pedestrian Information and AI Cameras

This paper presents the development of an AI-based system to assist visually impaired individuals in navigating safely. The system utilizes a wearable camera and advanced AI algorithms to identify and map "walkable areas" based on the presence of sighted pedestrians. This approach leverages human spatial awareness, providing a robust, real-time solution for navigating complex environments. Unlike conventional methods relying on obstacle detection or pre-defined maps, this system dynamically adapts to changing surroundings. It employs a pre-trained YOLO model for pedestrian recognition and an OAK-D Pro AI camera for real-time pedestrian detection and tracking. The system then generates a walkable area map based on the identified pedestrian movements displayed on a smartphone or tablet. This research addresses the limitations of traditional assistive navigation systems by leveraging pedestrian spatial awareness. The developed system empowers visually impaired individuals with greater independence and confidence in their daily lives by harnessing the power of AI and human spatial perception.