A Traffic Surveillance System in Real-Time to Detect and Classify Vehicles by Using Convolutional Neural Network

The vision-based surveillance systems are widely used in analysis traffic information due to their ease of installation and accuracy of their results. In this paper, an image processing system in real-time has been proposed to detect and classify the vehicles at intersections. This information can be used to estimate traffic density at intersections, and adjust the timing of traffic light for the next light cycle. The detection operation is performed by the background subtraction technique, the approximated median filter is used to extract and update the background, then the vehicles will be tracked in the detection area. After that the vehicle classification will be implemented by using the convolutional neural network (CNN). The system is applied to videos obtained by stationary cameras. The experiments demonstrate that this system is able to robustly detect and classify the vehicles.