Smart Vision: Enhancing Blind Mobility with Real-Time Object Detection, Proximity and Position Estimation

Visually challenged people use assistive technologies to navigate properly. However, existing mobility aids, such as smart glasses and robotic assistants, are often costly and out of reach for many users. This study introduces a sensor-free, computationally efficient assistive system that uniquely integrates bounding box-based distance estimation, position mapping using a 3 × 3 grid, and voice assistance for real-time navigation support. Unlike conventional approaches that depend on depth sensors or specialised hardware, the proposed system uses the YOLOv8s (You Only Look Once version 8 small) model to estimate object proximity directly from bounding box dimensions, thereby eliminating the need for additional sensors. A live webcam feed was processed to detect objects, estimate spatial positions, and compute distances to trigger proximity alerts. A voice assistance module delivers immediate auditory feedback, informing users of detected objects and potential obstacles at proximity. The system operates at 30 to 58 frames per second (FPS), and consistently delivers high-confidence detections ranging from 75% to 96%, ensuring reliable, low-latency performance across diverse environments. This integrated design enhances mobility, safety, and spatial perception for visually impaired users, without relying on expensive or high-powered hardware.