Title: Vision based vehicle tracking network and counting using deep learning model systems

Authors: K. Hemalakshmi; A. Muthukumaravel

Addresses: Department of Computer Science, Bharath Institute of Higher Education and Research, Chennai, Tamil Nadu, 600 073, India ' Faculty of Arts and Science, Bharath Institute of Higher Education and Research, Chennai, Tamil Nadu, 600 073, India

Abstract: Vehicle counting is a key component of the vehicle behaviourist approach and traffic incident detection for certain video surveillance systems. The accurate counting of vehicles in a variety of traffic conditions using deep learning algorithms and multi-object tracking systems is a popular area of research in the field of intelligent transportation. This research suggests a three-step vehicle identification, tracking, and counting process as a framework for video-based vehicle counting to estimate traffic flow. First, the Yolov3, Faster R-CNN, and SSD deep learning architectures are used to detect the vehicle, and the performances of each are compared. A modified DeepSORT algorithm tracks observed cars, and a picture shows their trajectory. In low-light and traffic conditions, a new vehicle counting system uses tracking data to count vehicles by type. Traffic figures are compared. The recommended method accurately recognises automobiles, tracks multiple objects, and detects with high precision and accuracy, according to experiments. This method meets real-time processing and vehicle counting needs. This study's method can count automobiles on difficult highways.

Keywords: vehicle detection; vehicle tracking network; vehicle counting; deep learning; traffic video; Yolov3; DeepSORT algorithm; detection-tracking-counting.

DOI: 10.1504/IJSSE.2025.148726

International Journal of System of Systems Engineering, 2025 Vol.15 No.4, pp.305 - 319

Received: 22 May 2023
Accepted: 24 Jul 2023
Published online: 22 Sep 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article