Artificial Intelligence-Based Real-Time Vision System for Autonomous Vehicles with Lane and Object Awareness

Main Article Content

Fardeen NB
Sameer NB

Abstract

Real-time perception systems for autonomous vehicles remain challenging due to the inherent trade-off between accuracy and computational efficiency. This paper presents DVNet-R, a real-time computer vision framework for autonomous vehicles incorporating optimized object detection and lane perception modules. The multi-stage architecture achieves 28.35 FPS on commodity hardware while maintaining high accuracy across diverse environmental conditions. Our system integrates a YOLOv8-based object detector with an 87.2% mAP@0.5 and a hybrid lane detection pipeline with adaptive region-of-interest selection achieving 93.5% accuracy. We contribute a novel adaptive processing technique that dynamically allocates computational resources based on scene complexity. Comprehensive evaluation across five environmental conditions demonstrates robust performance, particularly in variable lighting and weather scenarios. The modular design allows for targeted optimizations, as validated through our ablation studies. DVNet-R’s differential processing approach reduces computational overhead by 23% compared to baseline methods while maintaining competitive accuracy. This research advances the state of the art in efficient perception systems for autonomous driving.

Article Details

Section
Articles