PADS: A Reliable Pothole Detection System Using Machine Learning

Abstract

In recent years, with the popularity of private cars, a increasing number of people prefer autos as their way to travel. However, poor road conditions may cause damages to vehicles and have drawn great concern of governments all over the world. The extreme weather conditions, heavy traffics and low road quality are worsen the situation, making it is a challenging task to keep roads in good conditions. Therefore, frequent repairs are required to avoid damages to vehicles. In this paper, we propose a reliable pothole detection system using machine learning (PADS) to facilitate the road pothole detection and road conditions maintenance. The proposed system provides low latency in potholes detecting, thereby shortening the time for road maintainers to identify poor conditions roads. To make our system easy to deploy, we reduce monetary cots and simplify system architecture. To improve accuracy in potholes detection, we use \(K\_MEANS\) algorithm based on basic threshold algorithm. Our results display a plot of z-axis accelerations on one road and a pothole-marked map. At last, we show the pothole detection accuracy comparison between our algorithm and basic threshold algorithm.