Gaussian mixture modeling for video background substraction

Abstract

This project deals with Adaptive Background Modeling from an Image Sequence using Gaussian Mixture Modeling. It analyzes its system concept and algorithm steps. The Mixture of Gaussian which is proposed by Staufer and Grimson is a popular technique for modeling adaptive background in this research and many other researches. It deals with multimodal backgrounds that are exposed to complex environmental conditions, camera shakings and illumination changes .The Gaussian Mixture method deals with the detection of moving objects by modeling pixel grey level distribution along the time based on the learning and updating of background pixel distributions. This project employs the principle of the normal distribution and the basic concept of clustering using the Gaussian Mixture. The background subtraction techniques before GMM are discussed. The system is studied using a video example to study the parameters` update and the classification of pixels. Then, the experimental results show that the GMM technique demonstrates good segmentation and proves capable of using an adaptive method to accumulate data over time for constructing the background model.