Aiming at the problem of target tracking failure of kernel correlation filter( KCF) target tracking algorithm under the influence of occlusion factors
the original algorithm KCF is improved. Firstly
an occlusion decision is added to the original algorithm KCF. If no occlusion occurs
the original algorithm KCF is used for tracking; If occlusion occurs
the EM algorithm based on unscented rauch-tung-striebel( URTS) is used for target prediction. Secondly
the predicted target position is fed back to the original algorithm KCF. Finally
the video occlusion sequence is tracked and tested using the data of the OTB test library. The test results show that the tracking precision and success rate are improved compared to the original algorithm KCF.