MIL based visual object tracking with kernel and scale adaptation

作者：

Highlights：

• A fast online feature selection algorithm in MIL framework based on maximizing the classifier score.

• Introduces kernel trick on Haar-like features for target tracking.

• Explore use of Haar-features in half target space.

• A scaling strategy applied to get the actual target size in successive video frames.

摘要

Highlights•A fast online feature selection algorithm in MIL framework based on maximizing the classifier score.•Introduces kernel trick on Haar-like features for target tracking.•Explore use of Haar-features in half target space.•A scaling strategy applied to get the actual target size in successive video frames.

论文关键词：Computer vision,Visual tracking,Multiple instance learning,Weak classifier,Classifier score,Haar-features,Kernels

论文评审过程：Received 13 July 2016, Revised 21 January 2017, Accepted 23 January 2017, Available online 26 January 2017, Version of Record 3 February 2017.

论文官网地址：https://doi.org/10.1016/j.image.2017.01.007