一种基于SCP范数和结构稀疏的前景检测OA

Based on the robust principal component analysis(RPCA)model,aimed at the problems that the singular values are over-penalized and the spatial continuity of foreground is ignored in traditional RPCA models,a foreground detection model based on Schatten-Capped-P(SCP)norm and structure sparseness is proposed.The SCP norm was applied to constrain low-rank backgrounds which imposed different penalties on different singular values and could effectively avoid excessive punishment of singular values.The sparse foreground was constrained by the C(2,1)norm which took full use of the spatial continuity of the prospect.Experiments show that,compared with the five recent mainstream algorithms,the proposed model can effectively deal with the problems of excessive punishment of singular values and the spatial continuity of foreground.The proposed algorithm achieves the best average F-measure value.Compared with the suboptimal model,the average F-measure of the proposed model is increased by 44%in terms of per-formance.