An Adaptive Video-based Vehicle Detection, Classification, Counting, and Speed-measurement System for Real-time Traffic Data Collection

Intelligent Transportation System (ITS) is an integral part for efficiently and effectively managing road-transport network in metros and smart cities. ITS provides several important features including public transportation management, route information, safety and vehicle control, electronic timetable and payment system etc. In this paper, we have designed and developed an adaptive video-based vehicle detection, classification, counting, and speed-measurement tool using Java programming language and OpenCV for real-time traffic data collection. It can be used for traffic flow monitoring, planning, and controlling to manage transport network as part of implementing intelligent transport management system in smart cities. The proposed system can detect, classify, count, and measure the speed of vehicles that pass through on a particular road. It can extract traffic data in csv/xml format from real-time video and recorded video, and then send the data to the central data-server. The proposed system extracts image frames from video and apply a filter based on the user-defined threshold value. We have applied MOG2 background subtraction algorithm for subtracting background from the object, which separates foreground objects from the background in a sequence of image frames. The proposed system can detect, classify, and count vehicles of different types and size as a plug & play system. We have tested the proposed system at six locations under different traffic and environmental conditions in Dhaka city, which is the capital of Bangladesh. The overall average accuracy is above 80% for classifying all types of vehicles in Dhaka city.