Research On Parallel Compression Method For Multilevel River Linear Vector Data Considering Spatial Topological Relations

River data is the important data foundation of hydrological analysis, water eco-environment investigation and water quality monitoring, is the premise of constructing hydrological models, dynamic simulation of water pollution diffusion process and other related research. In the spatial data model of GIS, river data is often stored in the form of linear vector data. Large-scale and high-precision multilevel river vector data occupies a large storage space, and its speed of network transmission is slow. In recent years, with the wide application of digital map and high development of WebGIS, the problem is more and more prominent. Taking Chongqing’s multilevel river vector data as the basis, this paper studies the method of keeping the consistency of spatial topological relation in the compression of multilevel river vector data, and combining with the parallel computing, this paper studies the parallel compression method of multilevel river vector data to make the compression rationalized and high efficient.The main research work of this paper can be listed as follows:(1) As the four basis vector data compression algorithms, the principles of optical fence algorithm, vertical distance tolerance algorithm, angle tolerance algorithm and Douglas-Peucker algorithm are analyzed. Then this paper points out the outstanding advantages of Douglas-Peucker algorithm by analyzing the advantages and disadvantages of the four algorithms. Thus, DouglasPeucker algorithm is selected as the basic compression algorithm for multilevel river vector data.(2) This paper points out the inconsistency of spatial adjacency relations and the self-intersection or intersection in multilevel rivers when using the basic algorithm such as Douglas-Peucker algorithm to compress the multilevel river vector data, and analyzes the reasons causing the inconsistency of spatial topological relations. As to spatial adjacency relations in the multilevel rivers, firstly the adjacency nodes of multilevel rivers are extracted based on ArcGIS’s functions of topology analysis and network analysis, then Douglas-Peucker algorithm is improved to ensure it can reserve adjacency nodes by force. Thus the consistency of adjacency relations is kept in the compression of multilevel river vector data. As to self-intersectiion and intersection in the multilevel rivers, the search method is proposed based on scan line algorithm, then Douglas-Peucker algorithm is improved to eliminate the self-intersectiion and intersection by recovering the minutiae of multilevel rivers. On the basis of these, a serial compression algorithm for multilevel river vector data considering spatial topological relations is designed and implemented based on Douglas-Peucker algorithm.(3) Based on the serial algorithm, this paper studies some key issues in parallel computing. MPI is selected as parallel programming environment, SPMD is selected as parallel programming model, data-parallel model is selected as task assignment method and non-blocking point to point communication routine is selected as communication method, and then a parallel compression algorithm for multilevel river vector data is designed and implemented.(4) Based on Chongqing’s multilevel river vector data, this paper designs a replication experiment. Experimental data is compressed using the proposed parallel compression algorithm for multilevel river vector data. The experimental results are evaluated quantitative by length changing rate, tortuosity changing rate, relative displacement deviation, adjacency nodes retention rate and other indicators. Result shows that: when the threshold value is from 100 m to 1000 m, the length changing rate, tortuosity changing rate and relative displacement deviation reach an average of 0.607%、2.84%、0.506% using the proposed parallel compression algorithm. Compared with the Douglas-Peucker algorithm, the adjacency nodes retention rate increases by 35.15%, while the self-intersection rate and intersection rate reduces by 64.2% and 77.4%. The compression rate reaches 75.52% under a threshold value of 1000 m. These show that the parallel compression algorithm has good compression effectiveness. Compared with the serial compression algorithm, the speed-up ratio reaches an average of 1.755 using parallel compression algorithm with 2 nodes and 2.815 using parallel compression algorithm with 4 nodes, and it shows that the proposed parallel algorithm improves the compression efficiency of multilevel river vector data.