Path Optimization for Autonomous Driving Using Lane Detection in Python

Abstract

Autonomous driving depends on real-time perception and decision-making for safe navigation, and one of the central 
problems in that pipeline is path optimization, which requires accurate lane detection and tracking. This paper 
describes a computer vision approach to lane detection implemented in Python with OpenCV. The system processes 
video frames from multiple sources, applying grayscale conversion, Gaussian blurring, Canny edge detection and the 
Hough Line Transform to identify lane boundaries. A region of interest is defined to focus on the section of the frame 
where lanes actually appear, filtering out trees, signs and vehicles that would otherwise register as edges. The detected 
lanes are overlaid on the original frames, giving a visual representation of the optimized path. The system handles 
multiple video streams and was tested across different road conditions and lighting. The pipeline is deliberately built 
from classical operations rather than a trained model, which keeps it fast enough to run per-frame on modest 
hardware and keeps its failures interpretable: when detection breaks down, the stage responsible can be identified. 
The limitations of that choice are also examined, since a rule-based pipeline is sensitive to lighting, cannot handle 
sharply curved lanes well, and has no way to infer lane structure where markings are missing.