Adaptive Weighted Sampling Context-aware Correlation Filter Tracking Algorithm Based on Multi-feature Fusion

Aiming at the shortcomings of traditional correlation filter such as single feature and insufficient background information， an adaptive weighted sampling context aware correlation filtering algorithm based on multi feature fusion is proposed. Firstly， for gray image sequence， Histogram of Oriented （FHOG） feature， Local Binary Patterns （LBP） feature and gray feature are fused； for color image sequence， Histogram of Oriented （FHOG） feature， Local Binary Patterns（LBP） feature and ColorNaming （CN） feature are fused. Secondly， the high response region in the response graph is adaptively sampled， which is introduced into the filter training as negative samples， and the highest response region is given higher suppression weight. For the problem of target scale change， scale pool is introduced to estimate the scale. Finally， the algorithm is tested on OTB100 data set. Compared with the original algorithm， the tracking accuracy and success rate of the proposed algorithm are improved by 8.5% and 14.7% respectively， and compared with other mainstream algorithms， its effectiveness is verified .