Skip to Main Content
In this paper a technique for motion detection that exploits the Gaussian mixture models (GMM) and basic background subtraction (BBS) is proposed. For every frame, each pixel is modeled with almost K Gaussian distributions. All the existing GMM based techniques use a threshold to set a priori the number of Gaussians to represent the background. The proposed approach avoids setting this threshold. The results show the effectiveness of the novel approach on benchmarks test sets sequences.