Pothole Detection and Prediction using Sensors and
Machine Learning
Journal:
GRENZE International Journal of Engineering and Technology
Authors:
Nikhil K. Khaneja, Kalidindi Lakshmi Sanjana, Raunak Bhupal, Kala Bharatan
Volume:
6
Issue:
2
Grenze ID:
01.GIJET.6.2.1_1
Pages:
1-8
Abstract
In India, in metropolitan cities, a common problem seen is traffic jams or
slowdown on traffic due to potholes. Most of the road accidents are due to potholes. In this
paper, the authors have tried to ensure that the road surface quality and the potholes should
be monitored continuously and must be repaired as necessary. The best possible distribution
of resources could be attained when the real-time data is gathered from the mounted
sensory system and displayed and visualized in a way (Google Maps) which is
comprehensive and empirical. Different sensors would be mounted onto the public
transport vehicles. The method mentioned in this paper uses the ultrasonic sensor and the
accelerometer to send the data regarding the intensity of pothole straightaway to the cloud
data storage, along with the location coordinates of the potholes and the time stamp.
Different databases such as traffic density of the city, rain intensity of different areas is used
to train different machine learning models to predict the intensity of potholes at the
particular locations and get the best possible accuracy. The best-case accuracy attained was
35.5%. The pothole intensity data is visualized through a heat map application. This system
helps in reducing the frequency of traffic jams in any city of India. The method suggested in
the paper would help in reducing the number of accidents happening in India due to the
potholes.