A Real Time Traffic Analysis System using Computer Vision

In this paper, a vision-based real-time traffic analysis system is presented, which can analyze vehicles in traffic from a traffic video sequence. This paper discusses object detection, and tracking of objects in multiple video frames. The functionalities of traffic analysis using computer vision include vehicle speed estimation, traffic flow direction estimation, traffic density estimation and car colour determination. To detect objects in the traffic flow and to track objects Optical Flow Model and Kalman Filtering methods are used in this paper respectively. These algorithms are also used in determining the traffic density, vehicle speed and vehicle colour. Block Matching technique is used to determine the traffic flow estimation. Experimental analysis for colour estimation shows an accuracy of 85.71%. The results of this work culminates in object detection, object tracking, traffic density, vehicle speed, vehicle colour and traffic flow estimation which could be used for applications such as traffic control, security and safety both by government agencies and commercial organizations.