摘要
人体姿态跟踪是人类行为分析和识别的基础,具有广阔的应用范围。该文以人体的头肩部轮廓构成的Ω形状的曲线作为跟踪目标,利用多摄像机间的位置约束和形状约束,提出了一种基于多摄像机的人体朝向、位置和头肩部轮廓的联合跟踪算法。该算法将多视角头肩部轮廓跟踪和三维空间中的头肩部位置跟踪、朝向检测融合于同一框架中。真实场景实验表明该算法具有良好的鲁棒性和准确性。
Human posture tracking has long been of interest for its wide applications in human motion recognition and analysis.A tracking algorithm was developed using the Ω-shaped curve of the human head and shoulder as the tracking target based on the position and shape constraints among distinct views in a multi-camera environment.The algorithm integrates tracking and detection to simultaneously track the human posture and location in 3-D space as well as the head-shoulder contour.The robustness and efficacy of the approach is confirmed in challenging real-world scenarios.
出处
《清华大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2011年第7期966-971,共6页
Journal of Tsinghua University(Science and Technology)
基金
国家自然科学基金资助项目(60673189,60433030)
