【摘要】 轨迹聚类算法可以广泛地应用在交通管理中,利用轨迹聚类算法找出车辆轨迹热点区域对交通部门规划管理交通出行有重要指导意义。目前的轨迹聚类算法多以空间相似性进行度量,不能体现不同时间段的轨迹热点区域划分情况。针对上述问题,该文结合时间因素,提出了一种时空轨迹的热点区域提取算法。首先,对传统的密度峰值聚类算法进行了改进,考虑了计算密度的线性和非线性部分,改进了密度的计算方法;同时,改进了簇类中心的选取方法,能够自动地选取簇类中心;在此基础上,提出了聚类融合算法,过滤了不合适的聚类和多余聚类;最后利用DB检验量来检测提取效果。实验结果表明,相比于传统的聚类算法,本文算法能更有效地提取时空轨迹的热点区域。更多还原

【Abstract】 Trajectory clustering algorithm can be widely used in traffic management. Finding the vehicle trajectory hotspots by using trajectory clustering algorithm has important guiding significance for traffic planning and management of traffic travel. Current trajectory clustering algorithms are mostly measured by spatial similarity, which cannot reflect the division of trajectory hotspots in different time periods. In response to the above problems, this paper proposes a hotspot region extraction algorithm for spatio-temporal trajectory, combined with the factor of time. Firstly, the traditional density peak clustering algorithm and the density calculation method are improved by considering the linear and nonlinear parts of the calculated density. At the same time, the method of choosing cluster center is modified to enable it to automatically select the cluster center. On the basis of the above, we propose a clustering fusion algorithm to filter inappropriate clusters and redundant clusters and use the DB index to detect the division results. The experimental results show that our algorithm can extract the hot spots of spatio-temporal trajectories more effectively than the traditional clustering algorithms.更多还原