Multi-camera multi-target tracking method and device

By setting up independent and public databases in a multi-camera environment, using the pose estimation model and the improved DeepSORT model for target prediction and tracking, the migration problems of traditional algorithms and the occlusion loss information in a multi-camera environment are solved, and the accuracy and ID consistency of target tracking are improved.

CN119722751AActive Publication Date: 2025-03-28CHONGQING UNIV OF POSTS & TELECOMM
9 Cites 0 Cited by

Patent Information

Application Number
CN202411877167.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-12-19
Publication Date
2025-03-28
Estimated Expiration
2044-12-19

AI Technical Summary

Technical Problem

The traditional DeepSORT target tracking algorithm cannot be directly migrated in a multi-camera environment, and the accuracy rate is not high due to occlusion in a single-camera environment.

Method used

Multiple cameras are set up at different locations in the target area, each camera is independent of the database, and a public database is set up. The pose estimation model is used to predict the target, the improved DeepSORT model is used to match and track the target, and the bounding box, appearance characteristics and position feature information are extracted through the pose estimation model, and tracked with the Kalman filter and the Hungarian matching algorithm.

Benefits of technology

It effectively solves the target occlusion problem, improves the accuracy of target tracking results, and ensures ID consistency between multiple cameras.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention discloses a multi-camera multi-target tracking method and device, cameras are respectively and independently arranged at different positions of a target area, each camera is provided with an independent database and a public database, and the method comprises the following steps: acquiring a real-time video of each camera, and converting the real-time video into each frame of image data according to a time sequence; performing target prediction on each frame of image data by using the attitude estimation model, and extracting feature information of a to-be-tracked target according to a prediction result; and performing target matching and tracking on the feature information of the to-be-tracked target and information in the corresponding independent database or the public database by using the improved DeepSORT model to obtain a tracking result. According to the invention, the accuracy and robustness of pedestrian tracking in a multi-camera multi-target tracking system can be improved.
Need to check novelty before this filing date? Find Prior Art