A novel data decomposition method for rapid parallel processing of vector polygon rasterization

Chen ZHOU,Manchun LI,Zhenjie CHEN,Penghui JIANG,Dong CHEN

DOI: https://doi.org/10.11887/j.cn.201505004

2015-01-01

Abstract:According to the load balance problem of large-scale parallel vector polygon rasterization,a novel data decomposition method was proposed.Firstly,the number of polygon nodes or the number of polygons was employed to evaluate the amount of calculations of a subset.The spatial locations of decomposed lines were computed iteratively and the balanced calculations between decomposed subsets were guaranteed,so as to realize data decomposition and load balancing.Secondly,a binary-tree based fusion strategy was put forth to merge the polygons across multiple subsets.The proposed parallel algorithm was implemented under a multi-core CPU-based environment and multiple China land use datasets were employed.Experimental results show that the presented method can outperform conventional methods for different datasets and can achieve a higher speed-up ratio and good load balancing.Moreover,when dealing with a large-scale vector dataset,the number of polygonal nodes is more appropriate to be the metric to evaluate the calculation of a subset precisely.