Target tracking method and device, electronic equipment and storage medium

CN115345903BActive Publication Date: 2026-05-29SHENZHEN UNIV
3 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
SHENZHEN UNIV
Filing Date
2022-07-28
Publication Date
2026-05-29

AI Technical Summary

Technical Problem

Existing technologies struggle to achieve accurate target tracking in aerial photography scenarios, especially when the target size varies greatly, the shooting equipment moves rapidly, and the lighting conditions change drastically, resulting in insufficient robustness of target tracking methods.

Method used

Feature enhancement is achieved by employing a Siamese mutual attention module. Feature extraction networks are used to extract features from example and search images, and the feature set of the search image is enhanced using the Siamese mutual attention module. Combined with anchor box regression technology, robust, efficient and accurate tracking of the target is achieved.

Benefits of technology

It achieves robust, efficient, and accurate target tracking in complex scenarios, adapting to challenges such as large changes in target scale, rapid movement of shooting equipment, and drastic changes in lighting.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115345903B_ABST
    Figure CN115345903B_ABST
Patent Text Reader

Abstract

The application discloses a target tracking method and device, electronic equipment and storage medium, comprising: acquiring a to-be-detected video sequence and determining an example image and a search image; inputting the example image and the search image into a trained feature extraction network for processing to obtain a first feature map set and a second feature map set; inputting the first feature map set and the second feature map set into a trained twin mutual attention module for processing to obtain a reinforced second feature map set; obtaining regression features according to the first feature map set and the reinforced second feature map set; performing anchor frame regression according to the regression features to obtain the position of a to-be-detected target in the search image, and further determining the position of the to-be-detected target in the to-be-detected video sequence. The application strengthens the features through the twin mutual attention module, applies the information of the to-be-detected target to the extracted features, makes the adaptability of the reinforced second feature map set stronger, and realizes robust, efficient and accurate target tracking.
Need to check novelty before this filing date? Find Prior Art