Hybrid support vector machines for robust object tracking

作者：

Highlights：

• A novel hybrid formulation for object tracking is proposed.

• A novel regression model is presented to exploit more compact spatial constraint.

• Three different SVMs are used to build a specific tracker under above formulation.

• A distribution-based strategy is introduced to combine the hybrid SVMs.

• Experimental results demonstrate the state-of-the-art tracking performance.

摘要

Highlights•A novel hybrid formulation for object tracking is proposed.•A novel regression model is presented to exploit more compact spatial constraint.•Three different SVMs are used to build a specific tracker under above formulation.•A distribution-based strategy is introduced to combine the hybrid SVMs.•Experimental results demonstrate the state-of-the-art tracking performance.

论文关键词：Object tracking,SVM,Binary classification,Regression,One-class classification

论文评审过程：Received 25 July 2014, Revised 9 February 2015, Accepted 11 February 2015, Available online 25 February 2015.

论文官网地址：https://doi.org/10.1016/j.patcog.2015.02.008