Traffic analysis of avenues and intersections based on video surveillance from fixed video cameras

Abstract

Based on adaptive Gaussian mixture modelling this article presents the separation of foreground objects from frames of surveillance video taken at avenues and/or intersections. The paper also describes an approach for determining the lane fullness of a dedicated leg of an intersection. In order to give an accurate fullness measure the cast shadows that might be present in the segmented foregrounds must be eliminated. In this study the detection and removal of shadows have been carried out using the HSV color space. The simulations were carried out using the Camera 1 sequence from PETS 2001 database and a custom sequence recorded in TRNC-Famagusta. A new method for computing right and left lane fullness in each leg of the intersection has been proposed and values computed have been recorded on the bottom left corner of the frame under study. © 2009 IEEE.