© 2019, IJCSE All Rights Reserved 741 International Journal of Computer Sciences and Engineering Open Access Research Paper Vol.-7, Issue-3, March 2019 E-ISSN: 2347-2693 Semi-Geometrical approach to estimate the speed of the vehicle through a surveillance video stream A. Raj 1* , D. Dubey 2 , A. Mishra 3 , N. Chopda 4 , N.M. Borkar 5 ,V.S. Lande 6 , B.A. Neole 7 1,2,3,4,5,6,7 Dept. of Electronics and Communication Engineering, Shri Ramdeobaba College of Engineering and Management, Nagpur, Maharashtra, India Corresponding Author: raja@rknec.edu, Tel.: 7066060728 DOI: https://doi.org/10.26438/ijcse/v7i3.741748 | Available online at: www.ijcseonline.org Accepted: 22/Mar/2019, Published: 31/Mar/2019 Abstract—For the past few years reducing road accidents and controlling traffic by limiting the speed of vehicles has gained more importance. Most of the methods so far used are Doppler radar , IR or Laser sensor based speed calculation. All of them are very expensive and also their accuracy is not quite satisfactory. In this paper, a Camera-based Speed Calculation System(CSCS) is employed, CSCS uses image processing techniques and can process video stream in online or offline mode, CSCS has the ability to determine the speed with good accuracy but at relatively low cost. In this study, the acquired video is pre-processed to remove the redundant information, then foreground information is extracted from the video. After this noise and shadow are removed from the video. Moving vehicles are localized and centroid for them are found out. Region Of Interest(ROI) box was constructed for each lane. Speed is calculated with the help of Distance Speed Time formula by counting the number of frames taken by the vehicle to pass through the ROI box. A database in the form of the log file is created which contains vehicle speed, location(vehicle has passed from which CSCS system), time at which this speed was recorded and whether it has crossed the speed limit or not. CSCS was tested and has achieved satisfactory performance with an accuracy of 95.44%-99.64%. Keywords— Camera-based Speed Calculation System(CSCS), Background subtraction(BS), Localization, Centroid, Region of interest(ROI), Database, Automatic Number Plate Recognition(ANPR) system. I. INTRODUCTION According to the Indian Ministry of Road Transport & Highway, Global status report on road safety 2013 road accident statistics over 1,37,000 people were killed in road accidents in 2013 alone, which is more than the number of people killed in all our wars put together. Overspeeding is the major cause of car accidents that increases the risk of injury or death. According to the Governors Highway Safety Association Texas, overspeeding is a major factor in approximately a third of all traffic fatalities. It also plays a serious role in major injuries caused by car accidents. Therefore vehicle speed regulation is one of the most crucial carries out of the traffic laws. For determining the speed, Doppler radar was a reliable device as long as there was no other vehicle in the field of view. That's why the proposed study uses the Image Processing technique with the help of CSCS to accurately measure the speed of the approaching vehicle. Initially, the system was developed with a laptop and a mobile camera. The aim was to deploy the developed software into a compact system such as Raspberry Pi. CSCS can be integrated with the ANPR system to form a complete system.ANPR uses Optical Character Recognition (OCR) to extract the number plate characters of the vehicles and by doing so we can track the vehicles which are found breaking the speed limit. Rest of the paper is organised as follows: Section II presents the System Design and Implementation followed by Section III which presents the Experiment and Results. Section IV concludes this paper. II. SYSTEM DESIGN AND IMPLEMENTATION For accurate speed measurement camera placing becomes an important aspect. In our study camera is set up such that it records the front view of the approaching vehicle. The camera used in the study has a frame rate of 30fps. A frame is captured at every 33.3ms, the pixel size of the camera is 1.12 um Aperture and the focal length is f/1.7, 27.22mm. Perspective distortion on the acquired video was not considered in our study. Figure. 1 shows the block diagram for system architecture.