Real time vehicle classification and vehicle counting at intersection using deep learning techniques

Abstract

Fast suburbanization has resulted in more significant traffic backlogs demanding sophisticated traffic control solutions. This paper uses deep learning architecture to present a methodology for real-time vehicle classification and counting at intersections. Precise real-time vehicle classification and counting (RVCAC) at intersections are essential for efficient traffic management, especially in congested and heavy traffic mix conditions like those in Sri Lanka. Deep learning models deliver outstanding efficacy in object detection tasks compared to traditional machine learning models. Also, deep learning is a subcategory of machine learning. This paper explores a model that uses deep learning to classify and count vehicles around intersections. Our goal is to enhance accuracy by training a deep learning model based on a localized dataset that is specified for the Sri Lankan context. Real-time vehicle classification and counting, which are crucial for managing traffic conditions, detecting vehicle speed, identifying peak times, and more, have the potential to impact traffic management significantly.