Development of Law Enforcement System for Detection of Driver’s Seatbelt and Speed Estimation through Machine Vision

Abstract

Due the rapid growth in number of vehicles in Pakistan, traffic analysis has become a need of hour. Rate of accidents has gradually increased due to over-saturated roads, inexperienced drivers, unplanned movement and lack of respect to road safety rules. Vehicle safety condition is quite bad and necessities like airbags and reliable braking system are still considered as “luxuries” by vehicle manufacturers. Detection of vehicle speed and driver’s seatbelt, for safety concerns, have therefore become an essential requirement for traffic law enforcement agencies. This report presents law enforcement system for detection of vehicle speed and driver’s seatbelt based on machine vision. HAAR Cascade Classification is used for speed estimation and seatbelt detection. Vehicle tracking along the path in which vehicle enters the frame till the point when vehicle leaves the frame is required for speed estimation. Based on training images of our own data set collected at various locations, algorithm looks for possible seatbelt area in the upper body of driver’s area of vehicle and saves the detected image in a folder with pre-defined path. Assuming that vehicles are symmetric in shape, vehicle path is flat and monocular camera has negligible distortion, detection is done considering driving parameters in Pakistan. It has been found in results that detection, whether driver’s seatbelt is ON or not is about 87 and speed estimation is accurate by 90 percent.