Unmanned aerial vehicle video target detection method and system based on differential perception

By combining the differential sensing Transformer encoding module and the diffusion module, the problems of pseudo-motion interference and uneven target density in UAV videos are solved, achieving higher accuracy and robustness in target detection.

CN122454468AActive Publication Date: 2026-07-24QINGDAO UNIV OF SCI & TECH
5 Cites 0 Cited by

Patent Information

Application Number
CN202610942123.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-06-29
Publication Date
2026-07-24
Estimated Expiration
2046-06-29

AI Technical Summary

Technical Problem

Target detection in UAV videos faces challenges such as query-target mismatch caused by pseudo-motion interference and uneven target density distribution. Existing methods lack sufficient accuracy and robustness in complex environments.

Method used

A differential-aware Transformer encoding module is used to aggregate spatiotemporal features, and a differential-aware diffusion module is used to generate an adaptive target query. The differential features are used to suppress pseudo-motion and adapt to changes in target density.

Benefits of technology

It significantly improves the detection accuracy and robustness in UAV video scenarios, enabling stable and accurate target localization and identification in complex low signal-to-noise ratio environments, and enhances the generalization ability and adaptability of the detection model.

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

Abstract

The invention relates to the technical field of unmanned aerial vehicle target detection, in particular to an unmanned aerial vehicle video target detection method and system based on differential perception. The method comprises the steps of obtaining a to-be-detected unmanned aerial vehicle aerial video frame sequence, inputting the to-be-detected unmanned aerial vehicle aerial video frame sequence into a preset target detection model, and outputting a target detection result of each frame; the target detection model performs multi-scale feature extraction on each frame to obtain frame features; the differential sensing transform coding module is used for inputting a differential sensing transform coding module and outputting a differential feature and a coding feature; jointly inputting the differential feature, the coding feature and the initialized target query seed into a differential perception diffusion module to generate a target query aligned with the current frame; and performing feature decoding and target matching on the target query and coding features through a decoder to obtain a target detection result of each frame. According to the method, pseudo motion interference can be suppressed, the method adapts to dynamic change of target density, the problem of mismatching of query targets is solved, and the precision and robustness of unmanned aerial vehicle video target detection in a complex scene are improved.
Need to check novelty before this filing date? Find Prior Art