Fast-Moving Target Tracking Based on KCF with Motion Prediction

Tracking fast-moving targets with high accuracy is a challenging task in the field of target tracking. The speed of the algorithm should be taken into consideration when it is used in a real-time system. In this paper, we introduce motion prediction into the correlation filtering algorithms based on KCF and SAMF, which can achieve fast and accurate tracking of fast-moving targets. The target's motion characteristics are used to predict the position and scale information of the target in the detection frame, which greatly improves the performance of the correlation filtering algorithm. In addition, result evaluation and dynamic model update strategy are added to our algorithm to ensure that only the target's feature information is learned by the filter. Finally, the tracking result is refined using motion prediction and evaluation confidence for greater accuracy. The experiments demonstrate that our algorithm is more accurate and robust in tracking fast-moving targets and its speed is also greatly improved compared to SAMF.