An improved Gaussian Mixture Method based Background Subtraction Model for Moving Object Detection in Outdoor Scene

Detection of moving objects has become an essential step in video surveillance applications. Due to lack of automatic thresholding and self-adaptive updating of background model at pixel level, foreground-background separation is not correctly classified. In this research work, we have proposed a novel approach to detect moving objects (like person and vehicle) from static scene using single stationary camera. Firstly, we used statistical background model Gaussian Mixture Model (GMM) to generate the binary mask. At this stage, we have tuned GMM parameters to update the background model pixel wise. Then, blob analysis connected components labeling and morphology operations have been applied as post processing step to detect foreground efficiently. The proposed work was experimented on two benchmark datasets PET 2006 and Highway. The performance of the proposed approach is analyzed by calculating precision, recall, f1-score and accuracy. Experimental results reveal that the proposed method performs well as compared to popular Gaussian based and Non-Gaussian based methods