Object tracking system for traffic surveillance network

Abstract

An important aspect of traffic monitoring is the traffic surveillance The automatic identification of vehicle is mainly used for effective traffic management systems For traffic surveillance robustness and reliability are the basic components to show the effectiveness of the system There are many problems in detecting the vehicles Due to poor weather conditions and lighting conditions the performance of detection is degraded Thus this work presents an effective algorithms to detect and track the moving vehicles from a real video scenes captured by the stationary cameras in the highways Initially the moving objects are isolated from the background using a transform domain based Gaussian mixture modelling In this work the transform domain is chosen to reduce the computational time and complexity by replacing the pixel by pixel process with block level process which obviously reduces the execution time of the process In discrete cosine transform DCT the changes in the intensities are easily extracted which is suitable for different environmental conditions Generally the DCT blocks carry the valuable informations only in low frequency component which is less sensitive to external noises Also the modified discrete cosine transform is used to reduce the spatial redundancy. newline