基于视频数据和传递熵的鱼群信息传递网络构建  

Modeling for the information transfer network of fish school based on video data and information transfer entropy

作　　者：刘康生[1] 卢焕达[1,2] 于欣[2] 

机构地区：[1]浙江大学数学系,浙江杭州310027 [2]浙江大学宁波理工学院信息科学与工程学院,浙江宁波315100

出　　处：《控制理论与应用》2014年第7期858-863,共6页Control Theory & Applications

基　　金：国家自然科学基金资助项目(11231007;31201446);浙江省自然科学基金资助项目(Y6110751;LQ12C20006)

摘　　要：鱼群中的个体如何通过信息传递从而达到一致的群体运动状态,至今还没有刻画这一现象的统一数学模型.本文阐述了一种利用视频数据和传递熵构建鱼群中个体间信息传递网络的方法.首先用实验获取斑马鱼集群的视频数据,并采用计算机视觉跟踪的方法获取鱼群中每个个体的位置和运动速度,然后利用传递熵计算个体两两之间的信息传递关系,在此基础上构建了鱼群信息传递网络.通过网络分析,揭示了鱼群中个体间的信息交互个数与信息传播速度之间的关系,进一步发现了鱼群信息传递网络中的频繁子结构.本文提供了一种利用探测时间序列间因果关系建立鱼群信息传递网络的方法,为鱼群信息传递研究提供了一种新的思路.For how the individuals of fish school achieve the coherent locomotory status via information transfer,there is still no unified mathematical model.The paper describes a method to construct the information transfer network of fish school based on video data and transfer entropy.We first obtain the collective behavior video of zebra fish in the experiments and get the position and velocity of each individual using computer vision tracking method,and then calculate the transfer entropy between any two individuals to determine the information transfer relationship,finally build the information transfer network.The relationship between the number of information exchange and the speed of information transfer is revealed and the frequent substructures are detected through network analysis.We provide a new approach for fish school information transfer research by modeling the information transfer network based on the time series causality detection.

关 键 词：系统建模 集群行为 信息传递网络 频繁子图 

分 类 号：O157.5[理学—数学]