Moving object detection using self adaptive Gaussian Mixture Model for real time applications

Detection of moving objects is an important and difficult task in video surveillance systems. Background subtraction using Gaussian Mixture Model (GMM) is a widely used approach for moving object detection. Many improvements have been proposed over the GMM to accommodate various challenges experienced in video surveillance systems. This paper presents an improved GMM for moving object detection which models the intensity values of a pixel block using Gaussian components and uses dynamic learning rate. Due to the consideration of a pixel block instead of one pixel value, it takes 4 times less time than previously proposed methods and have performance almost similar to previous methods.