基于时间预测模型的高性能求交方法研究

Research on High-Performance Intersection Method Based on Time Prediction Model

多边形求交是GIS空间分析的基本操作之一,提高海量复杂多边形求交计算效率对大数据时代的空间分析具有重要意义。该文针对海量复杂多边形求交计算耗时长、数据与计算资源分配不均的问题,提出了基于时间预测模型的高性能并行求交方法:首先通过多级包围盒进行二次检测,减少求交对象的一对多关系;然后通过包围盒检测洞、岛的方法减少不参与计算的多边形,优化复杂多边形求交处理流程;最后,通过建立非线性回归模型度量求交运算的计算耗时,并据此优化并行计算负载,有效提高了海量复杂多边形求交效率。

Polygon intersection is one of the most basic operations of GIS spatial analysis.The computation of intersection of polygons with large number and complexity usually takes a long time.Therefore,it is of great significance for spatial analysis in the era of big data to improve the efficiency of intersection calculation of massive and complex polygons.In this paper,a high performance parallel intersection algorithm based on a time prediction model is proposed to solve the problem of time consuming and uneven distribution of data and computing resources.Firstly,secondary detection is carried out through multistage bounding box to reduce the one-to-many relation of the intersecting object.Then,the method of detecting holes and islands by bounding boxes is used to reduce the number of polygons that do not participate in the calculation and optimize the processing flow of intersection of complex polygons.Finally,a nonlinear regression model is established to measure the computational complexity of intersection computation,and the load of parallel computation is optimized accordingly.Experimental results show that the proposed method can effectively improve the efficiency of intersection calculation of massive and complex polygons.