针对现有聚类分析算法对平面空间时空轨迹进行聚类的不足，提出了基于区域的快速聚类方法。借助先验知识确定输入参数，使用符合现实情况的切比雪夫距离或曼哈顿距离度量点间距离以划分聚类簇，区分人的各个常驻地点；使用区域叠加的概念度量点的密度，确定簇的质心以获得每个常驻地点的中心坐标；在保证算法复杂度的前提下确定簇的边缘点，标注人在每个常驻地点的活动范围。相比现有算法，该算法更适用于时空轨迹数据分析。%In contrast to the deficiency in the existing clustering algorithms for spatio-temprol trajectory in planar space, put forward the fast clustering algorithm based on region, with the aid of prior knowledge to determine the input parameters, utilize Chebyshev distance or Manhattan distance which conform to the reality to measure the dis-tance between two points to partition cluster, therefore distinguish the resident place. Utilize the concept of region to measure the density of points, therefore determine the cluster center in order to get the center coordinate in each resident place. On the premise of guaranteeing the algorithm complexity to determine the bunch of edge points and denote people’ s activities scope in each resident place. Compared with the existing algorithms, the proposed algo-rithm is more suitable for spatio-temprol trajectory analysis.
展开▼
