基于动量动力学模型的行人关系分析  

Analysis of pedestrian relationships based on momentum dynamics model

作　　者：张旭光[1] 吴格非[1] 唐英干[1] 

机构地区：[1]燕山大学电气工程学院,秦皇岛066004

出　　处：《高技术通讯》2017年第3期245-253,共9页Chinese High Technology Letters

基　　金：国家自然科学基金(61271409;61273260);中国博士后科学基金(2012M510768;2013T60264);河北省自然科学基金(F2013203364)资助项目

摘　　要：进行了行人行为分析及行人关系研究。考虑到在机器视觉领域大多数研究仅关注于目标行为分类与识别,而目标间因果关系判别研究较少,且现有理论停留在心理学领域中的因果关系表达,提出了一种基于动量动力学模型的目标间因果关系识别方法,用于实现行人间因果关系的识别及量化计算。该方法利用Cam shift算法获得目标在视频中的位置并依据因果概念,构建动量动力模型,然后基于动量动力模型进行因果关系判别,最后计算因果值。根据因果值范围识别出视频行人间的三种因果关系:导致、促进和阻碍。实验表明,上述方法可以在视频监控条件下识别两运动行人间的因果关系。The pedestrian behavior analysis and the pedestrian relationship study were conducted. In consideration of the status that most of the current contributions in the field of computer vision focus on the classification and identification of target behaviors, while the distinguishment of causal relations between targets are less and the existing theories can only describe the causal relationship between two targets in the field of psychology, a method based on the momentum dynamics model was proposed to recognize the causal relationship between two pedestrians. The method is described below： Firstly, a Cam shift algorithm is used for obtaining the positions of targets in a video. Then, a momentum dynamics model is employed based on the concept of causality to distinguish the causal relationship between two pedestrians. Finally, the causal value is calculated. According to the range of causal value, causation, promotion and impediment, the three kinds of causal relationships between two pedestrians, can be gained. The experimental results show that the proposed method can identify the causal relationship between two pedestrians under the situation of video monitoring.

关 键 词：视频分析 行人行为分析 因果关系 社会力 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]