Orientation Method And Its Accuracy Evaluation For Irregular Aerial Image

Aerial photogrammetry is an important means for acquiring and updating the geographical information, is the main link to realize digital earth, is the main mode of mapping basic scale topographic map, has become a general operating mode of basic surveying and mapping in China. With the development of science and technology, photogrammetry has made significant progress based on the hardware technology. With the application of light-small flying platform, the operation mode has become more flexible, but faced with new challenge of instability posture at the same time. Based on the aerial triangulation, this paper focuses on the orientation of unconventional photogrammetry, investigated deeply the orientation method which is suitable to it, considering the new problem caused by irregular array and the large rotation angle, what is more, deeply exploring it’s positioning accuracy. The main research contents and innovative results are as follows:(1) Improved the fast simulation method of the unconventional aerial images.To enhance the efficiency of processing, this paper improves the image simulation method and proposes an inverse solution with panel as the processing unit to simulate the image. Firstly, select the DEM and DOM data of the corresponding region to be simulated. Then set the interior and exterior orientation elements based on the photographic conditions. Secondly, according to the characteristic of topography, image is divided into panel. And then iterate the elevation with the pixel of four corners of the panel to obtain the plane position of the corresponding ground point, what is more, establish the approximate affine transformation relations between panel and the ground grid and get the gray value from the DOM. Under a variety of photographic conditions, such a plain area and hilly area, according to the results of the simulated aerial images, the proposed method show t can simulate aerial images of all kinds of photographic conditions. Compared to the image simulation method with by pixel, the efficiency of processing is improved by about ten times.(2) Precision analysis of image matching in aerial photogrammetry. The precision statistics is carried out by the high precision matching of simulated images. The results reveal that when the geometric distortion among images is small, matching precision is better than1/6pixels, whereas matching accuracy is generally not more than0.4pixels. So the statistical results can be used as the precision estimation and checking index of image automatic measurement.(3) Aimed at the stereopair with large rotation angle, this paper proposes a rigid relative orientation method which uses singular value decomposition to constrain essential matrix in order to obtain the initial value of the unknown. Firstly, build the essential matrix between images, then use the traditional eight point method to linearly solve the elements of the essential matrix; secondly, considering the rank defect of the essential matrix, use singular value decomposition to descend rank in order to constrain essential matrix; then derive the initial value of relative orientation elements from constrained essential matrix; finally bring the initial value into rigid relative orientation model based on vertical parallax to process iteratively. According to the results of experiments, which use the simulated images, conventional aerial images and unconventional images respectively, the proposed relative orientation method can be applied to the large rotation stereopair and meets the precision request of reconstructing stereopair.(4) Aimed at the image model with big rotation angel, a new absolute orientation based on plane rotation is proposed. According to the rotation relationship between the normal vector of fitted plane to solve the three-dimensional rotation parameters directly is the kernel of this method. Firstly, an optimal plane is fitted according to the coordinates of model points in the image-space coordinate system and the object-space coordinate system respectively, subsequently the normal vector is calculated respectively. Then according to the spatial rotation relationship between the normal vectors, the initial values of angular elements are solved for absolute orientation. Finally, putting the above angular elements together with the scale factor which is calculated by the empirical formula as initial values, optimal solutions of the absolute orientation model in the rigorous model of spatial similarity transformation are obtained using the least square adjustment method. According to the results of experiments, which use the simulated images, real aerial images and close-range images respectively, the proposed absolute orientation method can be applied to the large rotation image model and totally meets the absolute orientation request in photogrammetric production.(5) To analysis the error propagation characteristics of aerial triangulation in each link and carry out statistics on precision estimation of unconventional image positioning. Through the block adjustment of simulated images in various conditions of photography, the formula of estimation of precision position for the unconventional aerial images is proposed.