现代电子技术2025，Vol.48Issue(19)：110-114,5.DOI:10.16652/j.issn.1004-373x.2025.19.018
动态场景下人体运动异常姿态跟踪
Abnormal posture tracking of human motion in dynamic scenes
武文杰 1张大志2
作者信息
- 1. 苏州科技大学,江苏 苏州 215009
- 2. 苏州大学,江苏 苏州 215009
- 折叠
摘要
Abstract
In dynamic scenes,human motion postures are complex and changeable,and there are interferences of complex lighting,background and occlusion,which bring certain challenges to the accurate tracking of abnormal postures.Kinect can obtain high-precision 3D human skeleton data in dynamic scenes in real time.The unscented transformation processing nonlinear system of unscented Kalman filter(UKF)can estimate the complex and changing human motion postures accurately.Therefore,the tracking method for abnormal postures of human motion in dynamic scenes is studied in combination with Kinect and UKF.The human motion data is collected by Kinect and a 3D skeleton model is generated,so as to construct a 3D motion database of human motion.By using the joint point data in this database with the UKF method,the posture vectors with interference noise in the dynamic scenes are predicted and updated to complete the whole posture tracking.The abnormal postures in the tracked posture vectors are determined by combining the Euclidean distance with the standard action posture vector.The results show that in the dynamic scene of illumination and motion changes,the proposed method can track different posture vectors of each joint effectively and identify abnormal postures accurately,so as to achieve the purpose of abnormal posture tracking.关键词
动态场景/人体动作/异常姿态跟踪/Kinect传感器/人体3D骨架/卡尔曼滤波/姿态向量/欧氏距离Key words
dynamic scene/human motion/abnormal attitude tracking/Kinect sensor/human 3D skeleton/Kalman filter/attitude vector/Euclidean distance分类
信息技术与安全科学引用本文复制引用
武文杰,张大志..动态场景下人体运动异常姿态跟踪[J].现代电子技术,2025,48(19):110-114,5.
