Occluded target tracking method, system and device based on autoregressive motion model

By combining an autoregressive motion model and a Kalman filter, the occlusion problem in multi-target tracking is solved, achieving accurate target tracking under occlusion conditions and improving the efficiency and accuracy of multi-target tracking.

CN117893571BActive Publication Date: 2026-07-03NANJING UNIV OF SCI & TECH
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
NANJING UNIV OF SCI & TECH
Filing Date
2024-01-16
Publication Date
2026-07-03

AI Technical Summary

Technical Problem

The serious occlusion problem in multi-target tracking makes it impossible for existing algorithms to accurately predict the target position, affecting the accuracy and efficiency of tracking results.

Method used

An autoregressive motion model is adopted. By acquiring the historical trajectory motion of the target and the trajectory motion of its neighbors, a trained embedded autoregressive model is used to extract the neighbor context information and predict the motion. The Kalman filter and K-means algorithm are then combined to reconstruct the motion trajectory of the occluded target.

Benefits of technology

It effectively addresses the problem of long-term multi-target occlusion, improves the efficiency and accuracy of multi-target tracking, and can accurately recover the motion trajectory of occluded targets under occlusion conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117893571B_ABST
    Figure CN117893571B_ABST
Patent Text Reader

Abstract

The application discloses a kind of based on autoregressive motion model's occluded target tracking method, system and equipment, it is related to multi-target tracking technical field, based on autoregressive motion model's occluded target tracking method by trained context information extraction network obtains the neighbor context information of occluded target, then the neighbor context information and the historical motion track of occluded target itself are input into motion prediction network to obtain predicted trajectory, finally, predicted trajectory is associated with remaining detection frame using KM algorithm, long time multi-target occlusion problem can be effectively coped with, and the efficiency and accuracy of multi-target tracking are improved.
Need to check novelty before this filing date? Find Prior Art