Research And Implementation Of Solution Method Of Bundle Aerotriangulation Adjustment

With the development of science and technology,a large number of emerging sensors have appeared one after another,and gradually replaced the traditional photogrammetry.UAV photography,vehicle camera photography,general hand-held camera photography and so on have become the main means to obtain geospatial information.Compared with traditional photogrammetry,they have the characteristics of low cost and high flexibility.In addition,they usually carry high-resolution digital cameras to obtain high-resolution images.However,with the continuous development of data acquisition methods and sources,the traditional method of bundle block adjustment has encountered unprecedented challenges in the face of memory consumption and low calculation efficiency brought by large amount of data.How to improve the adjustment efficiency of bundle block adjustment for large-scale data has become an inevitable problem.In order to solve the above problems,this paper focuses on two aspects: reducing the memory occupation and improving the adjustment efficiency,and proposes a fast and efficient solution.Aiming at the unique block structure of the bundle block adjustment,this paper proposes a processing method of large-scale sparse matrix block storage and solution,and seeks a fast solution method for the ill conditioned equation problem which often occurs in the adjustment process of the large-scale beam method area network.The main contents and innovations of this paper include:(1)For the problem that there are many unknowns in large-scale data,Schur complement of matrix is used to reduce the number of unknowns in bundle block adjustment.Large scale data will lead to large-scale sparse matrix in the process of bundle block adjustment,which cannot be stored and calculated by current computer technology.However,due to the regular sparse distribution of the spatial adjustment coefficient matrix in the area of bundle block adjustment,this paper adopts a special sparse matrix compression storage and solution method for its unique block structure.This method only stores and solves the non-zero elements,avoids the direct inversion of the coefficient matrix of normal equation,and effectively reduces the occupation of computer memory space.(2)Because the bundle block adjustment with large-scale data is often accompanied by the emergence of ill conditioned matrix,which will directly affect the stability of adjustment.In order to reduce the ill condition of the equation matrix of the bundle block adjustment,the conjugate gradient method is used to solve it.The results show that the conjugate gradient method can not only reduce the ill conditioned degree of linear equations,but also deal with sparse matrix effectively.(3)In order to further improve the adjustment efficiency of the conjugate gradient method for solving the modified equation,this paper introduces the preconditioned matrix and the inexact Newton method on the basis of the conjugate gradient method,so that it can improve the adjustment efficiency while ensuring the accuracy of the solution.(4)According to this method,the program is used to realize.Three groups of data of different orders of magnitude are selected for the experimental data,and the three groups of data are calculated and compared.On the one hand,the adjustment accuracy of the 3 groups of data meets the actual production needs,which proves the feasibility and effectiveness of the above methods;on the other hand,with the increase of the amount of data,this method can complete the calculation in a relatively short period of time,reflecting the high adjustment efficiency of this method.Compared with other methods,this method has obvious advantages in iterative convergence speed and point accuracy,which further proves that this method is the most effective metho d to solve the adjustment and modification equation of beam method.(5)At last,the idea of improving the accuracy of the grid leveling and the correlation between the inner and outer orientation elements is put forward.Following the development direction of big data era,the methods proposed in this paper are combined with related fields such as POS supported aerial triangulation,tilt photography,etc.