A joint method combining feature and deformation handling with classification model for object tracking

Object tracking is a widely researched topic with applications in event detection, surveillance and behavior analysis. There are three key steps in object tracking: feature extraction, deformation handling, and classification. In this paper, we present a joint method combining feature and deformation handling with classification model for object tracking. Multi-scale tracking map are obtained from multi-scale rectangle filters and sparse random measurement matrix. Then the map is put into a model combing feature and deformation handling. In the end, a BP net is used for classification. The cooperation is represented in the training process. Experiments on some publicly available benchmark video sequences demonstrate the advantages of the proposed algorithm over other approaches.