Traffic surveillance system for vehicles counting and speed estimation

Abstract

In this study, we present an algorithm to count vehicles and estimate their speed using cameras mounted on roads. The first step in the proposed algorithm is motion detection and then vehicles are detected after some threshold, morphological and contour operations. After detection process, vehicles can be counted and their speed can be measured. The vehicles counting part of the proposed algorithm was tested using 6 data sets which show different times and weather conditions and compared with the manual counting which is considered as ground truth. The vehicle speed measurement part of the proposed system was tested using 5 videos from five datasets available online. These data sets have a ground truth file for the actual speed of each vehicle. The actual speed was recorded by a high precision speed meter. The measured speeds were compared with the recorded ground truth values. Both parts of our system showed very good results. The overall counting accuracy was (90.8%) and the accuracy of speeds measurements that is within the acceptable error interval was (88.21%). The speed measurement was considered acceptable if it is within the (-5, +5 km/h).