Personnel tracking method and device based on state space model

By introducing state space model and YOLOv5 object detection model in multi-object tracking technology, combining Kalman filter and Hungarian algorithm, the accuracy and robustness of personnel tracking in high-density environments are solved, and more efficient tracking effect is achieved .

CN119314083AActive Publication Date: 2025-01-14ZHEJIANG UNIV
View PDF 4 Cites 2 Cited by

Patent Information

Application Number
CN202411423207.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-12
Publication Date
2025-01-14
Estimated Expiration
2044-10-12

AI Technical Summary

Technical Problem

Existing multi-objective tracking technologies have insufficient accuracy and robustness in personnel tracking in high-density environments, especially in the case of occlusion, lighting changes and nonlinear motion, resulting in inaccurate tracking results and errors in matching.

Method used

Using a personnel tracking method based on the state space model, combined with the YOLOv5 object detection model and Kalman filter, the target detection is optimized through GIoU loss, and the target matching is used using appearance loss and EIoU loss. The Hungarian algorithm is used for high confidence and low confidence groups. Target associations for , create tracking trajectories and remove unsuccessfully matched targets.

Benefits of technology

It improves the robustness and accuracy of personnel tracking, reduces missed detection and ID switching, is suitable for multi-objective tracking in complex environments, reduces dependence on manual verification, and improves detection performance and efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119314083A_ABST
    Figure CN119314083A_ABST
Patent Text Reader

Abstract

The invention discloses a personnel tracking method and device based on a state space model. The method comprises the following steps: acquiring monitoring video stream data; constructing a target detection model in combination with the state space model, and training the target detection model; transmitting the target to the trained target detection model frame by frame, detecting personnel existing in a picture to obtain a target detection result, and the category probability is the confidence coefficient of the target detection result; predicting the tracking result of the next frame of the tracking target by using a Kalman filter; performing Hungary algorithm matching on the high-confidence detection result and all tracking results based on the appearance loss and the EIoU loss; performing Hungary algorithm matching on a low-confidence detection result and an unsuccessfully matched tracking result based on EIoU loss; removing tracking target trajectories which are not successfully matched within a certain time; and creating a tracking trajectory for the high-confidence detection result which continuously appears and is not successfully matched, and incorporating the tracking trajectory into a set of tracking targets. According to the invention, the robustness of multi-target tracking can be improved.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Vehicle and pedestrian online detection and tracking method based on improved ByteTrack

    CN116682078A

  • Station dense pedestrian tracking system and method based on spatial weak clues

    CN118037774A

  • Polymer composition and production method therefor, combination, crosslinked object, and tire

    KR1020230140390A

  • Cross-video target tracking method and system, and electronic device and storage medium

    WO2023216572A1

Cited By

  • Multi-target vehicle tracking method, device and equipment under view angle of unmanned aerial vehicle and medium

    CN119887847A

  • Mamb-YOLO-based semiconductor laser chip damage detection method and system

    CN120543476A