A Novel HRBW-based approach to Mean shift algorithm for Target Tracking

摘要

To resolve the problem of localization errors in object tracking, which caused by the background pixels in an object model, a novel target tracking algorithm named Histogram Ration Background Weighted-based mean shift (HRBWBMS) is presented. In the proposed HRBWMS, unlike the standard mean shift, the target model is established based on object/background histogram log-likelihood ratio. A new weight transform method for target model based on object/background histogram log-likelihood ratio was introduced. The experimental results show that the proposed method not only accelerates the convergence, but also enhances anti-interference ability.