Street Eye: A Smart Vision-based System for Road Surface Monitoring Using Neural Networks

Abstract

Potholes on the road surface significantly impact road safety, vehicle performance, and passenger comfort. Traditional road inspection systems rely on manual surveying or sensor-based approaches, which are inefficient, expensive, and not suitable for large-scale, real-time road inspection. Recently, computer vision and deep learning technologies have made it possible to automatically analyze road conditions from visual data. In this paper, we introduce Street Eye, a smart, vision-based road-surface monitoring system that uses deep neural networks for automatic pothole detection. The proposed system uses a YOLO-based object detection algorithm to detect potholes from continuous video streams captured by a camera. For each detected pothole, geolocation data such as latitude, longitude, and timestamp are extracted and recorded in a centralized database. A location comparison function is designed to detect road users approaching potholes already detected, enabling real-time alerts. With the integration of real-time detection, geolocation data management, and alerting services, the proposed system is expected to improve road safety and facilitate intelligent transportation infrastructure by efficiently and cost-effectively monitoring road conditions.