| 25 | 0 | 35 |
| 下载次数 | 被引频次 | 阅读次数 |
轨迹相似性度量在轨迹数据化简、轨迹模式识别等应用场景中至关重要。现有轨迹相似性度量方法主要考虑目标的几何形态结构，在计算轨迹相似性时易丢失移动物体的运动、速度特征。本文提出一种融合多特征的轨迹相似性度量模型，综合考虑轨迹数据的空间、运动及速度特征，基于注意力图对比学习计算轨迹的相似性。首先，将轨迹表示为由节点和边构成的图数据结构，并设计多维度特征指标刻画轨迹数据；其次，通过设计图增强策略生成正负样本，从而丰富样本数据；然后，采用多跳图注意力网络编码器对正负样本进行特征编码，捕获节点间的关联信息，并利用无监督对比损失函数放大正负样本特征间的差异；最后，通过计算编码得到的多维特征向量间的欧氏距离度量轨迹间的相似性。采用GeoLife与深圳轨迹数据集进行模型训练与评估以验证模型的有效性，实验结果表明，模型准确率高达97.3%,明显优于基于几何维度的轨迹相似性度量方法。
Abstract：Measuring the similarity of trajectories plays a crucial role in applications such as trajectory data simplification and trajectory pattern recognition.Existing trajectory similarity metrics primarily focus on the geometric shape or structure of the target, which is often liable to the loss of important features related to motion and velocity of the moving objects when calculating trajectory similarity.To address this limitation, a new model frame for trajectory similarity measurement, called trajectory similarity-attention graph contrastive learning(TS-AGCL),is proposed.It considers not only the spatial and geometric features, but also the motion and velocity features of the trajectories.First, the model represents the trajectory as a graph data structure composed of nodes and edges, and designs multi-dimensional metrics to describe features of the trajectory from various dimensions.Secondly, to enrich the sample data, graph augmentation strategies are employed to generate both positive and negative samples from the input graph structure.Thirdly, an encoder using multi-hop graph attention networks is utilized to encode the features of both positive and negative samples, effectively capturing the relationships between nodes, and amplifying the differences between the features of positive and negative samples through a loss function based on unsupervised contrastive learning.Lastly, the similarity between trajectories is measured by computing the distance between the encoded multi-dimensional vectors representing trajectory features in the Euclidean space.To validate the effectiveness of the model, the Geolife and Shenzhen trajectory datasets are used for training and evaluating the model.Experimental results demonstrate that by integrating spatial distance, motion direction, speed, and other trajectory features, the model achieves an accuracy of 97.3% on the test set.The proposed model significantly outperforms conventional methods based on geometric features for measuring trajectory similarity.
[1] 杨敏，陈媛媛，金澄，等.保持移动速度特征的轨迹线化简方法[J].测绘学报，2017,46(12):2016-2023.
[2] 张永泽，达飞鹏.膨胀区域匹配和自适应轨迹管理策略的多目标跟踪方法[J].武汉大学学报(信息科学版),2024,49(4):572-581.
[3] 钱振，周侗，陶菲，等.基于聚类的AOI出入口位置提取方法研究[J].地理与地理信息科学，2020,36(2):1-7.
[4] 郭文峰，万义良，金瑞，等.基于共享单车轨迹的精细路网更新方法[J].地理与地理信息科学，2022,38(1):86-93.
[5] Su Han,Liu Shuncheng,Zheng Bolong,et al.A survey of trajectory distance measures and performance evaluation[J].The VLDB Journal,2020,29(1):3-32.
[6] Keogh E J,Pazzani M J.Scaling up dynamic time warping for datamining applications[C]//Proceedings of the 16th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining.New York:Association for Computing Machinery,2000:285-289.
[7] Kim J,Mahmassani H S.Spatial and temporal characterization of travel patterns in a traffic network using vehicle trajectories[J].Transportation Research Procedia,2015,9:164-184.
[8] Fréchet M M.Sur quelques points du calcul fonctionnel[J].Rendiconti del Circolo Matematico di Palermo (1884-1940),1906,22(1):1-72.
[9] 艾廷华.深度学习赋能地图制图的若干思考[J].测绘学报，2021,50(9):1170-1182.
[10] Li Xiucheng,Zhao Kaiqi,Cong Gao,et al.Deep representation learning for trajectory similarity computation[C]//2018 IEEE 34th International Conference on Data Engineering (ICDE).Paris:IEEE,2018:617-628.
[11] Chen Ziwen,Li Ke,Zhou Silin,et al.Towards robust trajectory similarity computation:representation-based spatio-temporal similarity quantification[J].World Wide Web,2023,26(4):1271-1294.
[12] Yao Di,Cong Gao,Zhang Chao,et al.Computing trajectory similarity in linear time:a generic seed-guided neural metric learning approach[C]//2019 IEEE 35th International Conference on Data Engineering (ICDE).Macao:IEEE,2019:1358-1369.
[13] Zhang Hanyuan,Zhang Xingyu,Jiang Qize,et al.Trajectory similarity learning with auxiliary supervision and optimal matching[C]//Proceedings of the 29th International Joint Conference on Artificial Intelligence.2020:3209-3215.
[14] Deng Liwei,Zhao Yan,Fu Zidan,et al.Efficient trajectory similarity computation with contrastive learning[C]//Proceedings of the 31st ACM International Conference on Information & Knowledge Management.New York:Association for Computing Machinery,2022:365-374.
[15] Han Peng,Wang Jin,Yao Di,et al.A graph-based approach for trajectory similarity computation in spatial networks[C]//Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining.New York:Association for Computing Machinery,2021:556-564.
[16] Yao Di,Hu Haonan,Du Lun,et al.TrajGAT:a graph-based long-term dependency modeling approach for trajectory similarity computation[C]//Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery & Data Mining.New York:Association for Computing Machinery,2022:2275-2285.
[17] 王海起，李留珂，陈海波.基于图结构的城市道路短时交通流量时空预测模型[J].地理与地理信息科学，2021,37(4):1-9.
[18] 时开鑫，陈跃红，张晓祥，等.基于图聚类神经网络的江西省山洪灾害危险性区划研究[J].地理与地理信息科学，2023,39(3):7-15.
[19] You Yuning,Chen Tianlong,Sui Yongduo,et al.Graph contrastive learning with augmentations[C]//Proceedings of the 34th International Conference on Neural Information Processing Systems.Red Hook:Curran Associates Inc.,2020:5812-5823.
[20] 师尚杰，李文德，闫浩文，等.图对比学习支撑下的矢量建筑物形状相似性度量[J].地球信息科学学报，2024,26(12):2659-2672.
[21] 彭瑞，赵文智，张立强，等.基于多尺度对比学习的弱监督遥感场景分类[J].地球信息科学学报，2022,24(7):1375-1390.
[22] Wang Guangtao,Ying R,Huang Jing,et al.Multi-hop attention graph neural networks[C]//Proceedings of the 30th International Joint Conference on Artificial Intelligence.2021:3089-3096.
[23] Chuang C Y,Hjelm R D,Wang Xin,et al.Robust contrastive learning against noisy views[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).New Orleans:IEEE,2022:16670-16681.
[24] Zheng Yu,Zhang Lizhu,Xie Xing,et al.Mining interesting locations and travel sequences from GPS trajectories[C]//Proceedings of the 18th International Conference on World Wide Web.New York:Association for Computing Machinery,2009:791-800.
[25] Zheng Yu,Li Quannan,Chen Yukun,et al.Understanding mobility based on GPS data[C]//Proceedings of the 10th International Conference on Ubiquitous Computing.New York:Association for Computing Machinery,2008:312-321.
[26] Zheng Yu,Xie Xing,Ma Weiying.GeoLife:a collaborative social networking service among user,location and trajectory[J].IEEE Data Engineer Bulletin,2010,33(2):32-39.
[27] Wang Guang,Chen Xiuyuan,Zhang Fan,et al.Experience:understanding long-term evolving patterns of shared electric vehicle networks[C]//The 25th Annual International Conference on Mobile Computing and Networking.New York:Association for Computing Machinery,2019:1-12.
基本信息:
中图分类号:P28;P208
引用信息:
[1]马鸿,李文德,闫浩文,等.基于注意力图对比学习的轨迹相似性度量[J].地理与地理信息科学,2026,42(03):9-18.
基金信息:
国家自然科学基金项目(42301513、42371463)
2026-06-02
2026-06-02
2026-06-02