摘要
提出了一种顾及空间物理约束的多密度网格聚类算法。该算法通过对障碍物和便利体两种物理约束的数据化处理,降低了聚类的复杂度。利用既有聚类数据又有障碍物的网格单元的二次分割方式来提高聚类精度。针对不同便利体对聚类影响的差异,引入便利度概念。用网格单元密度、单元间质心的曼哈顿距离和便利度三因素来构造判别函数,判别单元间的相似关系。理论分析和实验结果表明,在有任意形状物理约束的空间中,该算法能有效地对不同形状、大小和密度的数据集聚类。
A multi-density grid clustering algorithm to consider the physical constrains in space was proposed in this study. This algorithm can reduce the complexity of clustering through the data processing of the obstacles and facilitating body. The secondary segmentation approach of grid cells with both objective data and obstacles was used to improve the accuracy of clustering. The concept of convenience degree was introduced to target on the differences of the effect from the convenience to the clustering. A discriminant function, which was configured by inter-grid cell density, Manhattan distance between cell centroid and convenience, was used to discriminate the similarity relationship between cells. Theoretical analysis and experimental results showed that, in space with random shape of physical constrains,the algorithm could effectively cluster different shapes,sizes and density data.
出处
《测绘科学技术学报》
CSCD
北大核心
2014年第6期587-592,共6页
Journal of Geomatics Science and Technology
关键词
网格聚类
物理约束
单元
便利度
判别函数
grid clustering
physical constrains
cell
facilitator degree
discriminant function
