Traffic video surveillance: Vehicle detection and classification

Vehicle detection and classification is the most important and challenging stage of traffic surveillance using computer vision techniques. The videos captured using the closed circuit television (CCTV) cameras placed in roadsides or driveways are used for the surveillance. The surveillance system includes detection of moving vehicles, counting the number of vehicles and the classification of the detected vehicles. The main challenge of the computer vision technique is the real time applicability of the algorithms used. In this work a vehicle detection and classification algorithm which works in real time is proposed. The detection is carried out by the method of background subtraction where the background is modeled using the mixture of Gaussians and the detected vehicles are classified using the Artificial Neural Network (ANN) with a new set of features, Histograms of Oriented Gradients (HOG) and geometric measures of the vehicles. Experimental results shows that the proposed method with the new combination of features as training parameters for ANN give better result as compared to other popular algorithms.