ECE-VDTDA: A robust and computationally efficient collision avoidance system for driver assistance in foggy weather

Area of Science:

• Computer Vision
• Robotics
• Automotive Engineering

Background:

• Advanced Driver Assistance Systems (ADAS) and Collision Avoidance Systems (CAS) are crucial for modern vehicles, particularly for forward and rear-end collision warnings.
• Performance of ADAS and CAS is significantly degraded in adverse weather conditions like fog, posing safety risks.

Purpose of the Study:

• To propose an Efficient and Cost-Effective Vehicle Detection and Tracking with Driver Assistance (ECE-VDTDA) system to enhance ADAS and CAS performance in foggy weather.
• To develop and evaluate an optimized SimYOLO-V5s_WIOU vehicle detection algorithm and integrate it with advanced tracking algorithms for improved accuracy and efficiency.

Main Methods:

• Developed an optimized SimYOLO-V5s_WIOU vehicle detection algorithm incorporating SimSPPF and Wise Intersection Over Union (WIOU) loss.
• Utilized state-of-the-art Deep-SORT, Strong-SORT, and optimized Deep-SORT algorithms for vehicle tracking.
• Integrated detection and tracking modules with a driver assistance module for real-time collision warnings.

Main Results:

• The SimYOLO-V5s_WIOU algorithm demonstrated significant performance improvements, including a 17.45% increase in mAP50 on the foggy driving dataset compared to baseline YOLO-V5s.
• Achieved notable increases in multiclass mAP50, mAP50-95, F1 score, precision, and recall on the foggy cityscapes dataset.
• The ECE-VDTDA system enabled high-confidence vehicle tracking and demonstrated robustness and computational efficiency across multiple foggy weather datasets.

Conclusions:

• The proposed ECE-VDTDA system, powered by the optimized SimYOLO-V5s_WIOU algorithm, effectively enhances vehicle detection and tracking in foggy conditions.
• The system's driver assistance module provides timely collision warnings by estimating critical driving parameters, thereby improving road safety.
• The experimental results validate the robustness, computational efficiency, and practical applicability of the ECE-VDTDA system for autonomous driving applications in adverse weather.