Background Subtractionusing Online Matrix Factorization

  • K. Mounika, Ch. Rajendra Prasad


Object detection is a significant advance in numerous vision-based applications for programmed video assessment. Identifying a moving object in video frames for observation applications is a difficult issue. In conventional techniques, object recognition in a video is commonly completed by utilizing an object detector or background subtraction system. In present work, we come up with an efficient online background subtraction approach that could be applied to practical videos with assortment in the foreground and background. Unlike past methodologies that frequently from the foreground as LaplacianorGaussian conveyances, Our prototype the foreground with a set of Gaussian(MoG) distribution combinations for each frame, which is refreshed casing by frame on the web. In particular, the MoG method in every frame governed by the educated background/foreground information from the past frames. Also, the suggested models are defined as a fast- probabilistic MAP model, whatever the EM algorithm could solve voluntarily. We additionally join a relative change administrator in the proposed model that can familiarize itself with a wide scope of the background video changes the methodology increasingly dependable for camera developments.

Keywords: Low-rank matrix factorization, robust statistics, subspace learning, Background Subtraction

How to Cite
K. Mounika, Ch. Rajendra Prasad. (2020). Background Subtractionusing Online Matrix Factorization. International Journal of Advanced Science and Technology, 29(05), 9752 - 9758. Retrieved from