Improved KCF algorithm and its application to target lost prediction

In order to solve the tracking drift problem in the obscured environment and reduce the failure rate of target tracking in the obscured scene, this paper proposes a target loss warning mechanism in the obscured situation based on the traditional KCF algorithm, which uses the analysis of the 3D response map during the target tracking process, and uses the response maximum (Fmax) and the average value of the response between 2 adjacent frames to measure the tracking status of the target and determine whether the target has tracking drift. At the same time, the APCE evaluation criterion is used to reduce unnecessary model updates and increase the speed of computation. The simulation results demonstrate that the target loss warning mechanism can accurately warn the KCF algorithm when tracking drift occurs, and the tracking success rate and tracking accuracy can be improved by 9.5% and 5.3% respectively compared to the traditional target tracking algorithm in the occlusion scenario.