K-means++分区法在密集型测站中的应用研究

GAMIT软件在解算大型密集测站时一般需要进行分区处理,分区解算对结果的精度具有一定的影响。为了解决一般分区方法中长短基线同时存在而导致整网解算精度降低的问题,引入了K-means++算法和Hash算法实现分区,简称为K-means++分区法。首先利用K-means++算法对测站进行聚类,再利用Hash算法进行排序组合,这样能得到分布均匀的子网。文中采用整网解算结果作为标准值,分析区域分区法和K-means++分区法的基线长度、基线精度及三维坐标差,然后再将K-means++分区方法与间距分区法进行对比分析。实验结果表明,该方法比区域分区法精度更高,与现有的间距分区方法精度相持,且比间距分区法要稳定高效。 

Application of K-means++Partition Method in Intensive Stations

GAMIT software generally needs to perform partition processing when solving large-scale intensive stations. Partition resolution has a certain influence on the accuracy of the results. In order to solve the problem that the long-short baselines exist in the general partitioning method and the accuracy of the whole network solution is reduced, the K-means++algorithm and the Hash algorithm are introduced to implement partitioning, which is referred to as the K-means++partition method. First, using K-means++algorithm to cluster stations, and then using Hash algorithm to sort and combine, so that we can obtain an uniform distribution subnet. In this paper, the result of the whole network solution is used as the standard value. The baseline length, baseline accuracy and three-dimensional coordinate difference of the regional zoning method and the K-means++zoning method are analyzed. Then the K-means++partition method and the spacing zoning method are compared and analyzed. The experimental results show that this method is more accurate than the regional zoning method, and it is in line with the accuracy of the existing spacing zoning method, and is more stable and efficient than the zoning method.