Axle-Based Mapping Approach for Vehicle Classification in Heterogeneous Traffic

Abstract

Vehicle classification becomes a prerequisite for transportation planning and traffic facility design. Due to the increase in the variety of vehicle classes in heterogeneous traffic, collecting data becomes a tedious task. The current study proposes a methodology to obtain genuine and accurate vehicle class data from a mixed and large traffic data set using pneumatic tube-based traffic classifier equipment. Coupled with machine learning for data validation, this study employed supervised learning, relying on training the systems with real-time generated samples. The study aims to replicate the results of the device’s a vehicle classification scheme terminology provided by modifying Austroads94, for rigid vehicles and trailers, including six axle articulated vehicles (ARX) scheme of vehicle classification into nine different Indian standard motorized vehicle classes using an axle-based mapping approach. The algorithm proposes logistic regression for the preprocessing of the vehicle classification data set. The extensive sample size of over 7,200 different vehicles was collected for the analysis, among which, 50 individual vehicle class entries were used to train the data set. Passenger car unit values of the respective output vehicle class were used for model validation. The vehicle classification obtained from the proposed approach reached an accuracy score of 97%, predicting its utility over other measures. Wheelbase, which is the distance between a vehicle’s front and rear axles, was found to strongly correlate with vehicle classification. Accurate measurement of both vehicle speed and wheelbase can significantly improve the accuracy of vehicle classification. The current study emphasizes the crucial role of accuracy, convenience, and ease in data analysis by utilizing the automated traffic counter for data collection, addressing the challenge of classifying vehicles in large data sets in India’s diverse traffic conditions.