Relative Vehicle Velocity Estimation Using Monocular Video Stream

In the past few years, the intelligent driving systems have witnessed rapid development, either it is self-driving cars or driver assistant systems. All these systems are built around perceiving the environment of the vehicle and taking appropriate steps in the given context. Computer vision has been playing a significant role in reducing the number of costly sensors used to perceiving the environment. In the past, the velocity of the vehicle was major estimated using sensors. In this paper, we propose a data-based methodology to estimate the relative velocity of vehicles using monocular cameras hence omitting the need for costly sensors such as lidars. Our proposed methods achieve a low mean velocity square error of 1.806, for estimating the velocity of the vehicle in a real-time environment.