Vehicle Detection and Tracking at Night in Video Surveillance

This Many detection and tracking methods are able to detect and track vehicle motion reliably in the daytime. However, vehicle detection and tracking in video surveillance at night remain very important problems that the vehicle signatures have low contrast against the background. Traditional methods based on the analysis of the difference between successive frames and a background frame can not work. This paper presents a method for vehicle detection and tracking at night in video surveillance. The method uses Histograms of Oriented Gradients (HOG) features to extract features, and then uses Support Vector Machine (SVM) to recognize the object. In tracking phase, we use Kalman filter to track the object. As shown in experiments, the algorithm can exactly detect and track moving vehicles in video surveillance at night.