Research On Stitching Methods Of Urban Scene Images

With the continuous development of informatization and prevalence of the Internet, urban visualization research has become a hot topic, and panorama is the most direct and effective means of restoring the urban scenes. In real life, sometimes we have to stand at the center of a city community to look around the surrounding view, while sometimes we hope to appreciate exteriors of streetside architecture by strolling along the long street, which indicates the difference between a single-perspective panorama and a multi-perspective one.Basing on these different features of city scenes, this Paper takes urban scene images captured through ring shot and random photographing as main research object, conducts alternate stitching to ring-shot photos with current software technologies, outputs spherical panorama, and designs the single-perspective-based urban scene roaming system. The energy function is designed and improved on the basis of existing stitching techniques and algorithms, realizing the automatic stitching of multi-perspective panorama with a new means. Main content of the research includes the following three procedures:I. Ring shoot five groups of urban scene images with a common camera, conduct alternate stitching with Stitcher Unlimited2009, output the single-perspective spherical panorama, design the immersive360°panorama roaming system with Papervision3D Engine, realizing the method of creating a spherical panorama with non-professional equipment, and summarize, analyze both advantages and defects between single-perspective panorama and roaming system.II. Restore the moving track of multi-perspective camera and3D scene structure with SFM, optimize the Malkov random field, and modify its energy function with constraint condition so as to obtain the finest pixel region of each perspective image, realizing automatic stitching of multi-perspective panorama.III. Conduct automatic stitching to aerial-shot images of urban scenes with implementation method descried in Chapter III, change the weight of energy function, so as to keep the image seams off key buildings, realizing a panorama stitching of high resolution and wide scene with a small seam error.This Paper conducts stitching experiment mainly on aerial shot of Provence Community (Nanning) and images of its scene. The experimental results show that the stitching of panorama is smoother and more natural, with fine visual effect. The research is of certain practical and theoretical significance.Finally, this Paper draws conclusions to the work executed, and puts forward the direction for further research.