Accuracy Analysis Of Oblique Photography And Research On Optimization Methods

Oblique photography technology can combine efficient data acquisition equipment and oblique photography automation modeling software to deal with data collected faster and then output model.This new technology can not only directly reflect the appearance of the buildings but reduce the economy and time cost.In addition,the technology effectively guarantee reality and precision of the model.Image feature matching is a key link in oblique photography modeling.Both Pix4 D and Smart 3D adopt SIFT algorithm for feature matching in the process of space-three computation.However,because the SIFT algorithm takes a long time and further improves the accuracy of ariel triangulation,this paper proposes an improved feature matching algorithm based on Harris corner detection,combined with a number of UAV aerial photographs,and carries out experiments on MATLAB simulation platform.The results show that the improved algorithm proposed in this paper not only can achieve image matching faster,but also has better image matching effect.And it can solve the problem that SIFT algorithm has poor matching effect in regions with less feature points and similar terrain.As the oblique photography modeling technology is more advanced,the research on engineering accuracy analysis is less.Therefore,based on the project background of the Japanese-Russian Prison Museum in Lvshunkou District of Dalian,this paper designs two groups of precision analysis experiments of oblique photography model.This paper explores the more suitable engineering application scenarios of the two software by comparing the aerial triangulation accuracy and geometric accuracy of real three-dimensional model built by Pix4 D and Smart 3D post-processing software.In addition,comparing the effect of different angle image groups on the accuracy of true three-dimensional model,it is verified that the oblique angle image group can optimize the elevation accuracy and capture the texture information of the model more accurately.Finally,combined with several UAV aerial photographs in the project,the stability of the optimization algorithm is verified,and the matching effect is better than SIFT algorithm.The main research results are as follows of this paper are as follows.:(1)Firstly,this paper systematically analyzed the technical process and working principle of tilt photography,elaborated the principle of detection,description and matching in feature matching algorithm,and briefly summarized three constraints of feature matching stage.(2)Besides,this paper proposes an improved feature matching algorithm based on Harris corner detection.This paper design three sets of simulation experiments.The simulation results show that the optimal algorithm can extract the largest number of feature points,and the matching accuracy of the matching algorithm is the highest.It can solve the disadvantage of the SIFT algorithm in the condition of fewer feature points or similar environment.(3)Ultimately,this paper takes the Japanese-Russian Prison Museum in Lushunkou District of Dalian as the research area modeling,the oblique photographic modeling was carried out.The advantages and application scenarios of different software were illustrated by comparing the accuracy differences of real three-dimensional models built by Smart 3D and Pix4 D software.To study the difference of modeling accuracy between Left,Right and Mid images in Smart 3D modeling process which proved that the tilt image group can significantly improve the elevation accuracy of the model,while the vertical image group has better plane accuracy.Finally,combining with engineering application,the stability of the optimization algorithm under scaling,rotation and illumination transformation is verified,and the feature matching effect is better than SIFT algorithm.