Citation

Ghosh R. Int. J. Intell. Transp. Syst. Res. 2023; 21(2): 310-318.

Copyright

(Copyright © 2023, Holtzbrinck Springer Nature Publishing Group)

DOI

10.1007/s13177-023-00354-4

PMID

unavailable

Abstract

Modern intelligent transportation systems (ITS) require the automatic detection and tracking of vehicles moving on the road. Although several research works have been reported in this research domain, but a very few of them have focussed on the detection of moving vehicles. This article proposes an intelligent system based on an improved version of you only look once (YOLO) to detect and track the vehicles moving on the road. Machine learning (ML) based systems repurpose the classifiers to detect the vehicles, whereas the proposed improved version of YOLO model detects the vehicles as well as predicts the class probabilities in a single go from the input video frame in the form of an end-to-end system. The proposed improved version of YOLO contains 14 convolutional layers instead of 24 convolutional layers of conventional YOLO. The performance of this work has been evaluated on three different publicly available datasets, containing the videos of vehicles in varying weather conditions. The proposed vehicle detection system has provided 94.00%, 94.22%, and 95.67% average precisions on DAWN, CDNet 2014, and LISA 2010 datasets respectively. The detection results of the proposed system demonstrate that the proposed method outperforms the state-of-the-art methods.

Language: en

Keywords

Deep CNN; Modified YOLO; Moving vehicle detection