Vehicle Speed Measurement on Urban Roadways

Abstract

Worldwide, reckless driving is the root cause of numerous road accidents. In the past ten years, 4,73,084 traffic accidents have been reported in India. India has seen a significant increase in traffic because there is no way to control or monitor the speed of vehicles on the roads. The implementation of brand-new and cutting-edge speed enforcement technology is required to address this issue and reduce the number of fatalities caused by accidents. For the purpose of measuring vehicle speeds on urban roadways, we propose a video-based, non-intrusive system. To effectively locate vehicle license plates in motion-filled image regions, our system makes use of an innovative text detector and an improved motion detector. After that, distinct features on the license plate regions are chosen, tracked across multiple frames, and perspective distortion is corrected. The tracked features' trajectories are compared to known real-world measurements to determine vehicle speed. A data set with approximately five hours of videos recorded by a single inexpensive camera in a variety of weather conditions and ground truth speeds obtained by an inductive loop detector served as the test bed for the proposed system. Our data set can be used for research at no cost. Over 96.0 percent of the time, the measured speeds stay within the [-3,+2] km/h limit set by regulatory authorities in several nations, with an average error of -0.5 km/h. According to the authors, there are no other video-based systems that can match the results of an inductive loop detector. We also demonstrate that, with a precision of 0.93 and a recall of 0.87, our license plate detector outperforms both a well-known license plate detector and two other published state-of-the-art text detectors. The design of a device that estimates the speed of vehicles passing on highways from real-time video and keeps a log of the speed of vehicles tracked at a specific time is the goal of this project