<p>Modern driver assistance systems require reliable environmental understanding to ensure safe navigation under dynamic road conditions. Most industrial solutions rely on high-cost sensor fusion frameworks and computationally intensive deep learning architectures, limiting their accessibility in academic environments. This paper presents VIDAS (Vision- Based Intelligent Driver Assistance System), a modular real-time perception framework developed using accessible hardware and open-source technologies.</p>

<p>The proposed system integrates classical computer vision techniques with modern deep learning methods to perform lane detection, vehicle and pedestrian recognition, traffic signal classification, pothole detection, fog density estimation, night vision enhancement and turn-direction guidance. YOLOv8 is used for object detection, while Canny edge detection and Hough Transform are utilised for geometric lane extraction. A contrast- based visibility model is introduced for fog classification. In addition, a priority-based voice assistant module provides real- time audio alerts to improve driver awareness and safety.</p>

<p>The proposed architecture bridges theoretical concepts of intelligent driving and practical implementation while maintaining modularity, scalability, and computational efficiency suitable for academic research environments.</p>