An Efficient Vehicle Tracking Approach for Traffic Density Detection in Intelligent Transportation Systems Using Image Processing

In this paper, an efficient algorithm for traffic density detection in intelligent transportation systems is proposed. In the detection stage of this approach, the feature points extraction is used. Moreover, for grouping the feature points of each vehicle, background subtraction is used in areas closer to the camera and from the back view of the vehicles. In these areas, the newly arrived vehicles are recognized easily, and the features can be grouped at the same time. As the vehicles move away from the camera, the grouped feature points are tracked using the KLT algorithm. Accordingly, the vehicles are counted until even a feature of a vehicle exists in the area of interest, specified for traffic density detection. Therefore, even in an occlusion condition, if only one extracted feature of a vehicle exists, that vehicle will be detected. In addition to the accurate density detecting, the proposed method uses a larger area of the road of interest. Furthermore, due to the importance of real-time processing in traffic videos, it is provided by sampling the frames. Based on the proposed method, the traffic density can be detected with the accuracies of 98.9% and 97.8% in relatively light traffic and heavy traffic, respectively.