Research On Dense Matching Of Multi-View Oblique Images

Oblique aerial photography technology can apperceive the comprehensive information of the complex scene with the way of high precision and large range. It has great potential in the new generation of the city data infrastructure. As an effective supplement for conventional approximate vertical aerial photography, the datas that achieved by oblique aerial photography have been widely applied to the new city construction, Internet map services and the refined 3D city modeling etc.. So the processing theory and technology of multi-view oblique aerial datas has become the new hotspot in the field of photogrammetry and remote sensing.This paper expounds the acquisition technology of multi-view oblique images(such as multi-view oblique cameras and the oblique aerial photography) and the characteristics of the oblique aerial images; introduces the theories and methods of dense matching such as MPGC, MPM and PMVS etc., and deeply analyses the advantages and disadvantages of every method. The paper analyses the problems of PMVS in the dense matching of multi-view oblique images, which combines the practical need of three-dimensional modeling of cities to improve PMVS algorithm, and obtains one improved PMVS algorithm. The experiment shows that the algorithm in this paper can effectively improve the problems of PMVS, and provide a new method and technology for the dense matching of multi-view oblique images. The main contents of the paper are as follows:(1)the acquisition technology of multi-view oblique aerial imagesThe paper introduces the characteristics and the main technical parameters of typical oblique cameras at home and abroad, such as Pictometry, SWDC-5, TOPDC-5 etc., summarizes the development course of oblique aerial photography, and analyses the characteristics of oblique images and the advantages in refined 3D city modeling.(2)the theories and methods of dense matchingThe paper introduces some multi-view dense matching methods, such as MPGC, MPM, PMVS etc., and analyses the advantages and disadvantages of these methods from theories, accuracy. Thus, the paper chooses the Patch-based Multi-view Stereo algorithm(PMVS) that is more efficient and pithy to the research for dense matching of multi-view oblique aerial images.(3)improved PMVS algorithmThe paper analyses the disadvantages of PMVS algorithm in the dense matching of multi-view oblique images, and combines with the objects features in three-dimensional modeling of cities, makes use of elevation constrains, the selection of optimal object points and candidate images sorting strategies, it also uses a grid diffusion to substitute the original hexagonal-direction diffusion, and forms a algorithm suiting to the dense matching of multi-view oblique images. At last, the experiment verifies the feasibility and effectiveness of this improved algorithm.