基于单目相机的人物交互的识别方法、装置、设备及介质
By extracting and fusing human and object features using a monocular camera recognition method, and recovering motion trajectories using temporal information, the problem of low accuracy in human interaction recognition in existing technologies is solved, achieving efficient and stable human interaction recognition and 3D reconstruction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING UNIV OF POSTS & TELECOMM
- Filing Date
- 2026-04-29
- Publication Date
- 2026-07-17
AI Technical Summary
In existing technologies, human interaction recognition methods based on monocular cameras rely on human body trajectories to determine object trajectories during model prediction, which leads to stationary objects being misjudged as moving objects. Furthermore, when objects are occluded, their trajectories are easily distorted or lost, resulting in low recognition accuracy.
By acquiring multiple consecutive images, human and object features are extracted and fused to determine the temporal information of each feature. The temporal information is then used to determine the motion parameters of the human body and the object, ultimately identifying human interaction information. This avoids the interdependence between human body trajectory and object trajectory, and utilizes temporal continuity to recover the complete motion trajectory when the object is occluded.
It improves the accuracy and stability of human interaction recognition, reduces equipment costs and deployment complexity, is suitable for different scenarios and shooting angles, and enhances the stability and real-time performance of 3D reconstruction.
Smart Images
Figure CN122116424B_ABST