Research On Modeling Of DEM By Parallel Computing Technology Based On PVM

As a branch of DTM(Digital Terrain Model),DEM(Digital Elevation Model)is increasingly being used in many aspects of human's life. For example, DEM is being used in reconnaissance of bridge and road,design of building,realizing of GIS,analysis of martial terrain.Nowadays, how to turn out the DEM of a huge area fleetly is an important direction of this domain. In process of building the DEM of a huge area, much or even a great deal of data will be analyzed and calculated complexly. In a condition of limited calculational resource, the process of building the DEM of a huge area will be prolonged by the increased amount of data even finished without any result. A new method to make the DEM by parallel computing is analyzed and discussed detailedly in this paper. In the first place, after discussing and analyzing three traditional solutions to make the DEM, a new method based on the Algorithm of Making DEM Dividually(AMDD) and the Algorithm of Making DEM Point-by-Point(AMDPP) is proposed, called the Algorithm Making DEM Dividually and Point-by-Point(AMDDPP). AMDDPP is combined with the strongpoint of both AMDD's and AMDPP's. Through testing real data, it is testified that AMDDPP is a good algorithm and the more important point is AMDDPP can be disassembled to parallel algorithm easily. In the second place, a parallel computing system based on PVM is discussed. PVM is a software package can be used to build a virtual parallel computing system based on PC or workstation quickly and effectively. After analysing serial algorithm of AMDDPP detailedly, the AMDDPP is transformed to a parallel algorithm triumphantly in parallel level of task and transplanted in PVM successfully. The structure of the master and slave of the PVM was designed in this paper. Finally, a model used to analyse accelerative efficiency generally is put to action here to analyse the accelerative efficiency of the system based on PVM to making DEM, and the method of how to get better accelerative efficiency of the system is brought forward.