Structure-preserving learning method and apparatus for visual tracking using structure-preserving sp

文件类型：PDF文档
文件大小：1307K
The present invention provides a method and an apparatus for structure-preserving learning for video tracking using structure-preserving sparse learning, which are to robustly separate an object and a background in order to track an object in a video by providing an effective structure for maintaining a visual tracking method using maximum margin projection on the basis of l1-normalization optimization of sparse learning on the basis of selection of a corrected multi-tasking function. The method comprises the steps of : (A) designating a target function based on sparse learning on the basis of maximum margin projection (MMP)-based sparse subspace learning for robust separation between an object and a background; (B) generating a sparse cumulative set by integrating the designated discriminatory sparse learning-based objective function; and (C) generating a visual tracking method using maximum margin projection by performing optimization using an accelerated proximal gradient shrinkage method on the basis of the generated sparse cumulative set.