REAL-TIME VEHICLE IDENTIFICATION, TRACKING AND COUNTING SYSTEM USING YOLO AND DEEPSORT

Abstract

The count of motor vehicles on the road are continuously on the rise since industrial revolution took place. Due to this, quick and reliable vehicle detection, tracking and vehicle counting, on the road are required. However, the present approach focuses on counting the total number of vehicles without considering the direction and heterogeneity of vehicles. Therefore, there is a pressing need to design an efficient method to detect vehicle’s type, track, and count in up and down direction. Considering this in view, we design an efficient method to conquer the aforementioned task by utilizing deep learning methods, You Only Look Once (YOLO) and DeepSORT that helps to identify vehicle’s type, tracking the vehicle and estimate traffic density using YOLOv4 and DeepSORT. Accuracy metric is employed to find the effectiveness of the proposed model . Moreover, the accuracy metric of the presented framework is compared with the latest techniques which demonstrates that the presented framework performs better than the latest approaches: YOLOv3-tiny, YOLOv3, YOLOv4-tiny, DeepSORT by improving the detection accuracy by 12.16 %, 3.07%, and 7.69%, respectively and counting accuracy by 27.50%, 4.40%, and 11.00%, respectively.