Deep Learning Based Pothole Detection and Reporting System

Abstract

Humps and potholes are the foremost reasons for road accidents. It should be detected and informed to the other vehicle that is going to pass in that location leads to reduce accidents. To overcome this problem, in this paper, a novel road surface monitoring system is proposed for identifying the humps and potholes. The signals scattered from the ultrasonic sensor influenced to a large extent by the hump, and potholes of the road. Due to a reduction in the amplitude of the reflected signal, the above problem is hard to analyze. For real-time analysis, Kirchoff's theory has used. To overcome the limitations of Kirchoff's theory, Convolutional Neural Network-based Deep Learning (CNN-DL) has proposed for detecting the pothole and humps on the road. The location of the pothole has measured by a global positioning system (GPS) and updates the information to the control room. To prove the validity of the proposed method for estimating the potholes on the road, two other benchmark methods, namely, Kirchoff's theory, and k-nearest neighbor (KNN) are selected to validate the performance. The experiment results show that the CNN-DL is better than other methods for detecting pothole of the road at any kind.