Computer Vision based Vehicle Detection and Speed Estimation for Road Safety

EasyChair Preprint 7099

Abstract

Road safety is of immense significance, as travel is a basic necessity. According to data published by the World Health Organization (WHO), nearly 1.3 million people lose their lives due to accidents and road crashes. This research work focuses on deploying object detection and tracking algorithms to differentiate over-speeding vehicles in the traffic using Machine Learning concepts such as Computer Vision and Deep Learning.  A MobileNet SSD architecture trained on the MS COCO dataset for vehicle detection and identification was adopted. Subsequently, detected vehicles were tracked with respect to their centroids, and with the help of a simple velocity equation, the speed of the particular vehicle was registered. Moreover, the vehicles over-speeding would be captured along with their timestamp. The captured images could be shared over file hosting platforms such as dropbox, google drive, etc. authorities could make use of this system to bring orders on roadways through imposing fines on the offenders. Apparently, it can reduce accidents and road crashes.

Keyphrases: Centroid Tracking, computer vision, deep learning, object detection

@booklet{EasyChair:7099,
  author    = {Mohammed Ibrahim and B S Divya and Stephan Thompson},
  title     = {Computer Vision based Vehicle Detection and Speed Estimation for Road Safety},
  howpublished = {EasyChair Preprint 7099},
  year      = {EasyChair, 2021}}