Speeding Vehicle Detection in Unconstrained Environment

Abstract

In this paper a method is proposed to identify speeding vehicles in two phases, in the initial phase using YOLOV3 vehicles are detected, and in the later phase blur detection is used to identify whether the detected vehicle is speeding. We are able to achieve commendable results and accuracy of our results are directly dependent on the accuracy of YOLO object detection. To detect whether the vehicle is speeding Laplacian kernel is convolved against the gray intensity image and blur variance is computed. If the blur variance is in the order of 10^-4 and it is further decreasing across the neighbouring frames we can identify that the vehicle as speeding vehicle.