Vehicle Tracking and Speed Estimation Using Deep Sort
Giddaluri Bhanu Sekhar, M. Srilatha, Javvaji Srinivasulu, M. Bhargav Chowdary · 2022
In the developing countries like India, the vehicle riders are increasing day by day, where it also constitutes to increase in the number of accidents across the country. The current system which lacks of proper surveillance is the main cause for growth of accident rate each year. Proper surveillance in heavy traffic zones such as junctions might decrease the death rate caused by the most of the accidents by tracking the vehicles in real time. To overcome this problem, we need a system that analyze a footage of camera in a selected junction calculating motion and dimensions of vehicle using You Only Look Once (YOLO v4 which is implemented using Convolution Neural Network (CNN)) and tracking vehicles using Simple Online and Realtime Tracking (SORT) with deep correlation metric and finding their speeds frame by frame accurately with efficient computations. The system is an improvised version of current tracking architectures which is capable of classifying the vehicles and calculating their speeds.