数字地形分析中邻域统计型算法并行化方法及效率分析

Parallel Calculation and Efficiency Analysis for Neighborhood Statistic Algorithm in Digital Terrain Analysis

随着地理数据的来源不断扩展,分辨率不断提高,利用并行计算技术实现对海量地理数据处理已成为数字地形分析的研究热点。邻域统计型算法是一类常用算法,通过对一定分析窗口中的栅格值进行统计分析,以反映局部乃至区域地形特征。该文通过对邻域统计型串行算法解析,以地形起伏度算法为例,对该类算法的并行化方法进行了分析探讨,重点研究了数据划分策略和光圈效应处理策略。利用黄土高原DEM数据对算法的并行效率进行测试,结果显示,窗口大小和数据集规模对邻域统计型算法的并行效率有较大影响,当采用大数据集和大分析窗口时,算法呈现数据—计算密集型特点,其并行效率明显高于小数据集和小分析窗口的并行效率。

As the acquisition method for geographic data becomes more various and the resolution becomes higher, study on parallel calculation for massive spatial geography data has been a hot topic in digital terrain analysis. Statistic algorithms are in common use in digital terrain analysis,which could reflect the terrain characteristics in local or region scale by calculating statistics parameters within a certain analysis window. Based on the feature of neighborhood statistic serial algorithm, data partition strategy and the method for overcoming the halo phenomenon are discussed, and then the parallel method of such type algorithm is proposed. To verity the correctness and practicality of the parallel method, the experiment of relief algorithm is designed by using DEM data of loess plateau, the result shows that the window size and data size have a big impact on the parallel efficiency, the algorithm presents data and compute intensive characteristic when using big data with big window size and its parallel efficiency is much higher than that by using small data with a small window size.