An Improved Background Mixture Model for Robust Moving Object Segmentation

Gaussian Mixture Model (GMM) is one of the best models for modeling a background scene with gradual changes and repetitive motions. However, it fails in segmenting moving objects when the scene changes sharply. To handle this problem, an improved background modeling algorithm-Intelligent GMM (IGMM), which is inspired by the way human perceive the environment to tackle sharp changes in the scene, is proposed. In addition, each foreground pixel is relabeled according to its neighbourhood information in the binary foreground image to effectively reduce the number of False Negatives (FNs). The proposed method can make the GMM remember or forget what the scene has ever been during the learning and updating period. Experimental results show that it can help segmenting moving objects precisely when the scene changes sharply and improving the robustness of the scheme.