Research On Implementation Method Of Multi Source Fusion Velocity Field

In the development of the terrestrial reference frame,space geodetic technique has always been the main means to establish a high precision terrestrial reference frame.Based on the main factor that different space geodetic techniques have different advantages and disadvantages,how to design and realize the algorithm model of multi-technique fusion by drawing on each other's strengths has become one of the main research directions for the establishment of high-precision terrestrial reference frame.Since the velocity field is an important part of the terrestrial reference frame,it is of great significance in maintaining the reference frame and studying the plate motion.Based on the above reasons,this paper mainly studies the following contents:1.This paper studies the theory and algorithm of space geodetic measurements fusion,and fuses it with kalman filter algorithm,and deduces the algorithm model of space geodetic measurements fusion using kalman filter.2.The theory and algorithm of intra-technique combination has studied.The algorithm is implemented by software and the results are given.The results showed that GNSS had the highest accuracy of point position,followed by VLBI and SLR,and DORIS had the worst accuracy of point position,which was consistent with the conclusion in ITRF2014,which also proved the reliability of the software on the other hand.3.The theory and algorithm of inter-technique combination has studied.By introducing constraint vector,the coordinate and velocity of sites were collocated with local-ties constraints.The algorithm was implemented in software and the results of inter-technique combination were given.Compared with ITRF2014,the results calculated by the software are submillimeter level on GNSS site and millimeter level on VLBI,SLR and DORIS sites.The colocation sites have a velocity difference of millimeter level.This paper analyzes the results,obtains some conclusions and puts forward some solutions.4.The theory and its derivation under the condition of No-Net-Rotation are studied,and its realization method in kalman filter is studied,and the algorithm is implemented in the software to ensure that the multi-source fusion velocity field has No-Net-Rotation relative to the crust.5.The software for establish multi-source fusion velocity field using kalman filter is designed and developed.Cholesky decomposition,the optimization algorithm for matrix inversion,was researched and the software realized the algorithm.This paper studies the optimization algorithm of matrix multiplication--sparse matrix multiplication,gives the optimal storage mode of sparse matrix storage in COO format,compares the time consumed by using different matrix multiplication algorithms to solve matrix multiplication,and gives the optimization suggestions of matrix multiplication.6.The data of GNSS,SLR,DORIS and VLBI for three years of 2015,2016 and 2017 were processed at 862 sites,and the results of multi-source fusion velocity field calculated by the software were obtained.Comparison of the results with the ITRF2014 velocity field shows that: among 704 GNSS sites,there are 478 sites with velocity accuracy better than 1mm/a and 601 sites with velocity accuracy better than 5mm/a.Among 51 VLBI sites,15 sites had velocity accuracy better than 5mm/a,and 30 sites had velocity accuracy better than 1cm/a.Among the 40 SLR sites,only 1 site has a velocity accuracy better than 5mm/a,and 9 sites have a velocity accuracy better than 1cm/a.Among 67 DORIS sites,21 sites had velocity accuracy better than 5mm/a,and 35 sites had velocity accuracy better than 1cm/a.7.The results of the multi-source fusion velocity field are analyzed and discussed in this paper,the shortcomings are as follows:(1)Need to detect the sudden change of site coordinate and velocity and adopt corresponding strategies to deal with.(2)The short time span of data for the software to processing.The suggestions to solve the above problems are given and the future work is prospected.