基于关键帧加权特征融合的视频去重方法、系统及介质
By employing shot segmentation, keyframe weighted feature fusion, and shot closed-chain verification, the problem of insufficient accuracy in video deduplication in existing technologies has been solved. This enables effective identification of videos with the same source, videos with the same source at the beginning and end, and videos with partial overlap, thereby improving identification accuracy and recall efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENZHEN HUIDONG CREATIVE TECH CO LTD
- Filing Date
- 2026-05-07
- Publication Date
- 2026-07-17
AI Technical Summary
Existing video deduplication technologies struggle to accurately identify videos from the same source, videos with altered beginnings and ends, and partially overlapping videos. Their accuracy is particularly poor when dealing with video re-encoding, partial cropping, added visual interference, and platform packaging.
The video to be processed is segmented into shots, representative keyframes that meet the criteria of clarity and stability are selected, a subject kernel is generated and the perturbation envelope is expanded, non-subject interference areas are identified, keyframes are selected and features are extracted, and weighted fusion is performed in combination with the shot order to generate video-level candidate recall features. Shot matching edges are established and shot closed chains are constructed, and deduplication results are output.
It effectively distinguishes between the main content of a video and peripheral interference, improves the accuracy of identifying duplicate and nearly duplicate videos, reduces false positives, adapts to the rapid screening needs of large-scale video libraries, and enhances the accuracy of judgment.
Smart Images
Figure CN122157129B_ABST