基于距离变换的轨迹群组相似性度量

Similarity measurement of trajectory group based on distance transformation

从海量轨迹中挖掘移动对象的时空运动规律一直广受关注。轨迹数据的聚类、异常检测、多尺度概化等,都是以轨迹数据的相似性度量为基础的。相较于单一轨迹的相似性,轨迹群组的相似性受到尺度、轨迹数据的多维特征等多种因素的影响,缺乏广泛共识的度量方法。面向轨迹群组相似性度量的需求,本文提出了基于距离变换的轨迹骨架线提取方法,以实现轨迹群组的骨架线提取;并利用最长公共子序列算法开展轨迹群组骨架线的相似性度量;最后,基于GeoLife数据集对本文方法进行了验证。结果表明,基于距离变换的方法能够较好地顾及轨迹数据的多种形态特征。这可为轨迹群组相似性度量提供新思路。

With the widespread adoption of mobile positioning technology and the accumulation of trajectory data,trajectory data mining has emerged as a highly prominent research area.Clustering,outlier detection,and multiscale generalization of trajectory data are all rely on measuring the similarity of trajectory data.Among these approaches,measuring the similarity among trajectory groups has become a crucial aspect of trajectory analysis.Accurate quantification of trajectory group similarity is essential for uncovering concealed patterns and regularities within trajectory data.Therefore,exploring effective methods for measuring trajectory group similarity holds paramount importance in fully realizing the potential value of trajectory data.The purpose of this study is to investigate an efficient approach for measuring the similarity of trajectory groups from a grid-based perspective,utilizing the concept of distance transformation in map algebra.Through empirical research conducted on the GeoLife dataset,this study aims to validate the feasibility of the proposed method and derive conclusions to guide trajectory data mining and processing.In this study,a distance transformation-based method is adopted to measure trajectory group similarity.Firstly,the impact of different line widths values(L1)on the extraction of overlapping fractures and gaps is assessed,to observe their influence on feature extraction.Subsequently,different distance transformation values(L2)are employed to determine the scope of dilation,enabling the extraction of skeleton lines that adequately describe the overall distribution of trajectory groups.Finally,the similarity of skeleton lines at various scales is evaluated to assess trajectory group similarity and compare the effectiveness of different trajectory processing algorithms.The experimental results emphasize the significance of selecting appropriate line widths and distance transformation values for extracting skeleton lines of trajectory groups.Smaller line widths and distance transformation values help retain detailed features of trajectory groups,while larger values are more suitable for capturing the overall distribution.Additionally,the scale of skeleton lines is a pivotal factor influencing their similarity.Skeleton lines at the same scale exhibit higher similarity,whereas those at different scales exhibit lower similarity.The distance transformation-based method provides a means to not only quantify the similarity among trajectory groups but also capture the overall characteristics of extensive trajectory datasets.The choice of line widths and distance transformation values,along with the discrepancies in skeleton line scales,significantly impact the measurements of trajectory group similarity.Therefore,when processing and mining trajectory data,it is imperative to select appropriate parameters and methodologies based on specific application contexts and objectives.This study introduces a novel approach for measuring trajectory group similarity based on distance transformation,offering fresh insights and methodologies for trajectory data mining.Accurate measurement of trajectory group similarity facilitates the discovery of concealed patterns and regularities within trajectory data,providing essential guidance for trajectory data processing and analysis.Furthermore,this method can optimize the design of trajectory processing algorithms,thereby enhancing the efficiency and effectiveness of trajectory data mining.