A Structural Feature Based Automatic Vehicle Classification System at Toll Plaza

Abstract

Most of the existing toll collection systems in vehicle toll plazas in India are manual in nature. Automation of toll collection systems at toll plazas will make the system a lot faster and fraud-free. The primary task for building such a system is to classify the vehicles arriving at toll plazas because accordingly the amount of toll varies. Most of the existing works in this regard have focused on tracking and detecting of on-road vehicles, but very few of them tried to classify the vehicles. This article presents a novel machine learning based approach to detect vehicles arriving in toll plazas along with their types or classes. In this approach, various structural features are extracted from each vehicle before feeding those features to different classifiers. An exhaustive experiment has been performed on a large self-generated dataset using five different classifiers - Gaussian naive Bayes, Multinomial naive Bayes, Logistic regression, Random forest and Support Vector Classifier (SVC). An encouraging accuracy of 96.15% is obtained from the present system.