摘要
对交通监控中运动目标的轨迹距离计算和聚类方法进行了改进。在轨迹距离计算中,引入目标的空间坐标、运动速度、运动方向和尺寸4个参数,以提高聚类时对不同位置、不同速度、不同方向和不同尺寸运动目标的轨迹的区分能力;针对交通目标运动轨迹比较规律的特点,采用基于统计的方法对K均值的轨迹聚类算法进行初始化,从而可以自适应的确定聚类数目K值和聚类初始中心。在真实场景下,验证了算法的有效性和适用性。
An improved distance calculation and clustering algorithm on trajectories of moving objects in traffic surveillance scene is proposed.First,when calculating the distance between trajectories,four parameters,such as space coordinates,velocity,direction and size of the target are introduced,in order to improve the capability of distinguishing different objects' trajectories which located in different area,with different speeds,different directions,and different sizes of the objects.Second,In view of the characteristic that the trajectories of traffic targets are disciplinarian,the K means clustering algorithm is initialized by a statistical method,to solve the problems of determining the number of clusters and selecting the initial centers.Finally,the results of the experiments in real scene have confirmed the validity of the algorithm.
出处
《计算机工程与设计》
CSCD
北大核心
2012年第6期2417-2422,2427,共7页
Computer Engineering and Design
基金
中国博士后科学基金项目(20100471838)
陕西省自然科学基金项目(2010JM8014)
关键词
交通监控
轨迹分析
距离计算
聚类
行为理解
traffic surveillance
trajectories analysis
distance calculation
cluster
behavior understanding