Vehicle Detection and Tracking Methods

---------------------------------------------------------------------***-----------------------------------------------------------------------Abstract This paper presents a scheme for vehicle detection and tracking and categorization from pavement CCTV in traffic surveillance. The system counts vehicles and separate them into four categories: car, van, bus and motorbike. A new backdrop Gaussian Mixture Model and dark elimination method have been used to deal with sudden enlightenment changes and camera shaking. A Kalman filter tracks a vehicle to enable categorization by majority voting over more than a few successive frames, and a level set method has been used to purify the foreground blob. Wideranging experiments with real world data have been undertaken to assess system concert. The best routine results from training a SVM (Support Vector Machine) using a combination of a vehicle outline and intensity-based HOG features extracted next background subtraction, classifying forefront blobs with mass voting. The vehicle monitoring system is evaluated on a new dataset together on Italian motorways which is provided with fairly accurate ground truth obtained from laser scans. For a broad assortment of distances, the remember and exactitude of detection for cars are outstanding. Information for vehicles are also informed. The dataset with the ground truth is through communal.