DeepBus: Machine learning based real time pothole detection system for smart transportation using IoT

Road related accidents have always been a nuisance to drivers and pedestrians alike. Every year countless accidents and deaths occur due to potholes which could have been preventable if there had been a prior warning or if the civic authorities were able to repair these potholes in time. This paper proposes a machine learning based pothole detection system called DeepBus for real time identification of surface irregularities on roads using Internet of Things (IoT). DeepBus uses IoT sensors to detect potholes in real time while an end user is driving vehicles on the road. The location of these potholes would be available on a centrally hosted map which can be accessed by both end users and civic authorities. Thus, it would serve as a warning system to all users as well as a database of potholes with thier locations to the authorities for quick repair and action. We have compared the performance of various machine learning models (Logistic Regression, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Naive Bayes, Decision Tree, Random Forest and Ensemble Voting) based on different parameters (Accuracy, F-score, Precision and Recall) and identified that Random Forest is the best model for pothole detection.