IoT and Cloud Based Automated Pothole Detection Model Using Extreme Gradient Boosting with Texture Descriptors

Abstract

One of crucial activity related to road monitoring and maintenance is the occurrence of potholes. These potholes are also be major reason of road accidents, damaging of vehicles, discomfort of passenger journey and extensive in terms of time and cost. But, identification of potholes can significantly alleviate the aforementioned issues. Other side, the Internet of Things (IoT) plays a crucial role in different applications, and provides viable and state of art solutions for variety of problems. Hence, the aim of this work is to develop a real time automated pothole detection model to detect the potholes in asphalt roads based on IoT devices.The proposed model comprises of three main components such as collection of pothole data and labeling, image pre-processing and texture feature extraction, and extreme gradient boosting (XGBoost) algorithm. The potholes data on asphalt road is collected by three IoT sensors such as accelerometer, ultrasonic sensor, and GPS and further, the collected data is transmitted on cloud via Wi-Fi module. The texture features are extracted using Gaussian steerable and median filters. The extreme gradient boosting (XGBoost) classifier is adopted for prediction task. The simulation results showed that proposed XGBoost model obtains higher accuracy, recall, precision and F1-score rates as 94.56, 97.41, 96.40, and 96.90 respectively using 10-cross fold validation method.