一种运动特征提取方法、人体姿态估计方法、人体网络重建方法、设备及介质

By dividing video sequences into segments and performing adaptive pose pooling and cross-attention modeling, the boundary discontinuity problem caused by segmentation in video 3D human pose estimation and mesh reconstruction is solved, improving the continuity and accuracy of 3D pose estimation and mesh reconstruction.

CN122244472BActive Publication Date: 2026-07-17PEKING UNIV SHENZHEN GRADUATE SCHOOL
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
PEKING UNIV SHENZHEN GRADUATE SCHOOL
Filing Date
2026-05-22
Publication Date
2026-07-17

Smart Images

  • Figure CN122244472B_ABST
    Figure CN122244472B_ABST
Patent Text Reader

Abstract

本申请公开了一种运动特征提取方法、人体姿态估计方法、人体网络重建方法、设备及介质，所述方法包括将视频序列划分为多个视频片段；对视频片段进行二维人体姿态估计，以获取视频片段的二维人体姿态和第一图像特征；基于二维人体姿态对所述第一图像特征进行自适应姿态池化处理以得到第二图像特征，并基于二维人体姿态确定时序运动特征；对第二图像特征和所述时序运动特征进行时序一致性建模，以得到具有全局时序约束的运动特征。本申请通过在三维姿态映射阶段引入时序一致性建模，对视频片段内与视频片段间的人体姿态特征进行时序一致性建模，显式建模了人体姿态在时间维度上的依赖关系，缓解因视频分段处理带来的边界不连续问题。
Need to check novelty before this filing date? Find Prior Art