Road Feature Detection for Advance Driver Assistance System Using Deep Learning

Area of Science:

• Computer Vision
• Artificial Intelligence
• Road Safety Engineering

Background:

• Road accidents result in numerous injuries and fatalities annually.
• Driver inattentiveness to road features like vehicles, pedestrians, and signs is a major contributing factor.
• Timely awareness of road features can significantly reduce accident rates.

Purpose of the Study:

• To propose a computer vision-based system for detecting and recognizing traffic elements and signs.
• To enhance driver awareness and contribute to the development of self-driving cars.
• To establish a benchmark for improving traffic safety and enabling advanced automotive technologies.

Main Methods:

• A custom, real-world roadside dataset was collected under diverse environmental conditions.
• Two deep learning models, YOLOv7 and Faster RCNN, were trained on the annotated dataset.
• The models were evaluated for their ability to detect road features.

Main Results:

• YOLOv7 achieved a mean Average Precision (mAP) of 87.20%, and Faster RCNN achieved 75.64%.
• Both models demonstrated class accuracies exceeding 98.80%.
• The results represent a state-of-the-art performance for the task.

Conclusions:

• The proposed computer vision models offer a robust solution for real-time road feature detection.
• This technology serves as a foundation for developing advanced driver assistance systems (ADAS) and autonomous vehicles.
• The study highlights the potential of AI in significantly enhancing road safety and transportation.