Background Subtraction Based On Robust Principle Component Analysis And Its Implementation Using CUDA

In this paper,I propose an online background subtraction model based on robust principal component analysis(RPCA)named ORPCAI,and describe its GPU implementation and optimization.There are four parts in this paper:introduction,theoretical basis,details of ORPCAI and GPU implementation.In the first part,the paper mainly introduces the concept,significance and status of background extraction.The second part starts from the basic steps and methods of background extraction,then goes to the development and evolution from the background extraction algorithm based on principal component analysis to background extraction algorithm based on robust principal component analysis.In order to find a fast and robust background extraction model that could update itself immediately when a new frame is given and can be GPU-accelerated,I modified the basic robust PCA model and then found the optimization method and solution of my new model.And I also find that after adding the initialization in my ORPCAI model,it shows a better performance in several situations.At last part,I introduced the overall framework of my algorithm,the details of the algorithms uesd by ORPCAI.To use CUDA to make ORPCAI run efficiently on Nvidia GPU.I introduced several technologies like texture memory,CUDA linear algebra library,parallel computing in the algorithm optimization part.And the implementation and the usage of these technologies is also described in this section.