Real-Time Visual Navigation Framework for Edge-Level Systems Using Classical Vision Algorithms

Abstract

This paper presents a novel real-time navigation assistance system that enhances driver awareness by visually embedding directional guidance directly into the roadway scene. Instead of displaying instructions on a separate map interface, our system renders navigational cues - such as arrows - within the actual road view captured by a forward-facing camera. These cues are aligned with the detected lane geometry, enabling drivers to follow directions more intuitively without shifting focus away from the road. The system relies on lightweight computer vision techniques using OpenCV, avoiding the need for complex neural networks or high-performance hardware. It operates efficiently on minimal CPU resources and requires only a simple camera. Real-time integration with Google Maps provides up-to-date route information, which is seamlessly transformed into visual overlays projected within the driving lane. This design significantly improves driver focus and reaction time, particularly at higher speeds or in unfamiliar environments. By bridging map data with on-road visual guidance, our method offers a practical step forward in intelligent driver assistance systems. The approach supports future extensions such as windshield-projected interfaces, but its current implementation already demonstrates how smart, vision-based systems can make navigation safer, more accessible, and more human-centered. © 2025 IEEE.