Adaptive Tracking Algorithm for Multiperson Motion Targets Based on Local Feature Similarity

Abstract

The traditional multitarget tracking method has many points of concern. In view of the problems of high target recognition switching rate and high target trajectory false alarm rate in complex scenes and the problem of target loss caused by video tracking in existing video tracking algorithms, a new method based on video tracking is proposed: a multitarget tracking algorithm for local feature similarity, a strong target motion maneuvering or rapid deformation of asymmetric rigid targets. The algorithm builds a depth metric model, which can predict and track the temporal features of the target trajectory frame appearance features and motion features at the same time, which makes the extracted target features more discriminative and reduces the target recognition switching rate. At the same time, the adaptive model tracking algorithm can adjust the model in real time according to the clarity of the target area, effectively ensure the accuracy of the target tracking model, effectively aggregate the characteristics of the trajectory frame, and reduce the false alarm rate. The experimental results show that combining the multiperson moving target adaptive tracking algorithm based on local feature similarity into the DSST model can improve the average accuracy and success rate of the DSST model and has good stability.