Application Of Mpi Parallel Technology In Surveying And Mapping Data Processing

With the development of science and technology,the means of acquiring spatial data are becoming more and more abundant,and the types of surveying and mapping data are increasing as well as,the amount of data is exploding.To deal with massive data quickly and in real-time has become a hot research topic in the field of surveying and mapping.Multi-core and multi-processor computers are powerful tools for humans to solve large-scale scientific computation.As the continuous development of parallel technology,parallel systems,parallel programming environments,parallel processing algorithms,etc.,which are compatible with parallel computers,are also continually being improved.Among them,MPI(Message Passing Interface),as a representative of the message passing programming model in the field of parallel computing,has been widely applied to various disciplines and fields.In this paper,MPI parallel technology is applied to the data processing of surveying and mapping,and the research is mainly carried out from the following three directions:(a)Grid of gravity quantities parallel computing in the field of physical geodesy;(b)Image segmentation and parallel matching in the field of photogrammetry and remote sensing;(c)Parallel registration of multiple point cloud data in the laser point cloud field.The main research work of this paper is as follows:(1)Introduced the parallel computer system,compared the three parallel programming environments of message passing,shared-memory and data parallel,analyzed the parallel mode of MPI,and summarized the communication mode,functional characteristics and applicable scope of the calling interface commonly used by MPI.(2)To solve the time-consuming problem of computing large-scale grid gravity quantities by spherical harmonic synthesis formula,firstly analyzing the spherical harmonic synthesis formula,in order to avoid repeated calculation of the recursive coefficients of cos m? 、 sin m? and Legendre functions(ALFs)in the traditional algorithm,this paper proposed the method of calling pre-storage array,and then combined with MPI parallel technology to further improve computational efficiency.The test was carried out on DELL server and Tianhe-2 supercomputer respectively.The experimental results show that the MPI parallel technology can greatly improve the computational efficiency and greatly reduce the calculation time;The way of calling pre-storage coefficient can significantly improve the computational efficiency on the server with ordinary performance,but it will slow down the calculation speed on the supercomputer platform with excellent calculation performance.(3)SIFT(Scale Invariant Feature Transformation)algorithm has high matching accuracy,but poor real-time performance.In this paper,MPI is used to process image blocks in parallel.Aiming at the problem of cracks caused by image block matching,an overlapping segmentation method for image is proposed.The SIFT algorithm can be divided into two parts: feature extraction and feature matching,in which feature extraction takes up the majority of the time of whole algorithm,while the feature matching takes less time.Based on this feature,two parallel schemes are set up in the experiment: Scheme 1 extracts features in parallel,and collects features for serial matching;Scheme 2 extracts features in parallel,and then collects and broadcasts the features of one image to all processes for parallel matching.The experimental results show that the parallel processing of image blocks with MPI can significantly speed up the SIFT matching speed;The communication time of scheme 2 is more than that of scheme 1,but the feature matching time is reduced,and the overall efficiency is better than scheme 1.(4)ICP(Iterative Closest Point)algorithm is usually used for precise registration of point clouds.When ICP algorithm is used for pairwise registration of multiple point clouds,the process takes a long time.In this paper,local transformation matrix is introduced as transition matrix,and a series of point clouds are processed in parallel by MPI.The experimental results show that parallel processing of multiple point clouds by MPI can significantly speed up the registration and reduce the execution time.The text type point cloud is slow to read,and the parallel reading of multiple point clouds by MPI can greatly reduce the time consuming.