Study On Large-Format Image Acquisition And Super-Resolution Reconstruction Based On Close-Range Rotating-Camera

With the development of computer vision and digital photogrammetry,in recent years close range photogrammetry has been widely recognized and used in industry engineering filed such as environmental perception and road recognition of driver-less cars,3D modeling of smart city,dynamic monitoring and reverse engineering of aircraft carrier,large aircraft and industrial sheet metal and deformation monitoring of large-format precision engineering like building construction and hydraulic engineering.While the small size and low resolution of images caused by digital camera's narrow field angle and large light sensing units have reduced the stereoscopic visual range and measuring accuracy of close-range photogrammetry and are out of the requirement for precise close-range photogrammetry in high-precision industry engineering filed,so that the development of digital close-range photogrammetry is restricted.Targeting at the rotating-camera and focusing on the image capturing of high resolution close shot from wide field of view,this thesis managed to conduct studies on relevant theories and methods based on the main purpose of improving the application precision of close-range photogrammetry.The main achievements of the research are as follows.(1)According to the need of precise close-range photogrammetry on the ground to the image from wide field of view,a rotating-camera equipped with total station and industrial camera has been designed to elaborate--the relationship between rotation angle,overlapping of images,effective images and virtual image.The geometrical relationship's effect of large-format images on precision was deeply analyzed.Meanwhile,the mathematical relation between the largest rotation angle of rotating-camera and the change rate of resolution ratio and the relationship between the rotation angle of adjacent images and overlapping degree of photos were inferred,so that it optimizes the collection method of rotate Image and provides the image capturing of close shot from wide field of view with the systematic and optimized strategy of using rotating-camera.(2)Considering the uncertainty of images due to the movement of total station in the process of rotate shooting,the author has built a seamless splice model of large-format images which takes position offset of the camera into account based on the deep analysis of geometrical relationship between projected images fromlarge-format equivalent center and rotating images and the reliability of the model has been proved through the experiment.It has effectively enlarged the coverage area of images from single exposure station,eliminated the deficiency that the field angle of digital camera is small and supplemented the theories on precise close-range photogrammetry.(3)Aimed at the problem that the resolution ratio of digital camera is low,an in-depth study on super-resolution reconstruction of single image has been deeply conducted from the perspective of digital camera's incomplete sampling cognition and compressed sensing theory.By introducing compressed sensing increment-dimensional method,using dependency of high and low resolution images and combining learning method and and nonlinear reconstruction algorithms of compressed sensing,the author has created a super-resolution reconstruction method based on compressed sensing and pseudo-dictionary,so that the resolution of digital image and the reliability of reconstruction algorithms are enhanced and the information out of cut-off frequency is effectively recovered.Besides,the influence of the surface features and resolution ratio on reconstruction precision has been analyzed so as to improve the low resolution ratio of digital camera and conclude theoretical methods to improve the measurement precision of precise close-range photogrammetry.(4)The super-resolution reconstruction method based on compressed sensing and pseudo-dictionary has been applied in the three-dimensional reconstruction of Yanshan Park in Xuzhou City and super-resolution reconstruction method from large field of view used by rotating-camera has been applied in fine three-dimensional reconstruction of urban buildings.The feasibility of the proposed project and the reliability of the model are finally verified.