Road Pothole Detection Using Smartphone Sensors

Abstract

Pothole-related accidents are a significant problem in India, resulting in numerous yearly injuries and fatalities. Researchers have explored multiple ways to detect potholes on roads. However, the collection and analysis of the data present significant challenges, including data quality, accuracy, and processing time. A system has been proposed in this paper that can detect potholes on the road using a smartphone’s built-in sensors. Several parameters, such as acceleration, rotation angle and rotation speed, have been measured on and around the X, Y and Z axes using the Android smartphone’s inertial sensors. A Random Forest (RF) based machine learning (ML) model has been trained and tested on collected data for the proposed pothole detection system. It is observed that data quality is critical in determining machine learning models' effectiveness. Our experimental results show that the RF model is highly accurate in terms of accuracy, precision, recall, F1-score and AUC (area under the curve), which are 91.85%, 0.9180, 0.9182, 0.9181 and 92.318, respectively.