Road Surveillance using Gaussian Mixture Model for Birth and Clutter Events in Object Tracking

Abstract

In video Surveillance for real time images, particularly, when applied for vehicle tracking in roads, complexity arises due to the fact that multiple objects or vehicles appera or disappear from the scene. The modeling of a road is a multi- target environment, where multiple targets are present in the scene. The appearance and disappearance of the targets are modelled by Gaussian Mixture Model (GMM). The model developed was used by Probability Hypothesis Density (PHD) filter. The PHD filter utilizes the contextual information so that occluded targets can be identified. The tracks for entered object, hidden and then appearing object can be extracted from the video images.