Using High Resolution Satellite Images in the Bundle Adjustment of Aerial Images' Block to Reduce the Necessity of Ground Control Points

Introduction: Despite the emerging developed methods in photogrammetric engineering, aerial photogrammetry based on its traditional instructions is still known as a conventional method for generating spatial information over large regions. Aerial triangulation is a crucial step in preparing aerial images for extracting spatial information. Bundle adjustment is the most common method of performing aerial triangulation, in which exterior orientation parameters of images are estimated through Ground Control Points (GCPs), tie points, and other auxiliary observations. In photogrammetric engineering, by reducing the necessity for ground control points in the aerial triangulation process, strategic goals such as achieving direct georeferencing and reducing operational costs have been pursued. Integrating various spatial data sources to facilitate the aerial triangulation process is known as a traditional approach in photogrammetric engineering. So far, the contribution of navigation data, maps, geometrical constraints of manmade and natural features, and ortho-images alongside GCPs has been used for this purpose. In this research, the idea of using High Resolution Satellite Images (HRSI) along with their refined Rational Polynomial Coefficients (RPCs) to participate in aerial triangulation has been proposed. This was done using tie points measured between HRSI and aerial images. The uniformity in georeferencing accuracy of HRSIs has been the motivation for using such data sources to prevent systematic deviations in aerial triangulation within the spatial resolution of satellite images.

Materials and Methods: The proposed method introduces tie points between aerial and satellite images as new observations in the bundle adjustment. The rational functions associated with the HRSI produce the condition equations for these tie points. Before using the initial RPCs published along with HRSI, their preparation process, including geometric refinement and changing their ground coordinate systems, is performed. In the bundle adjustment process, the coefficients of the refined rational functions are held constant; their role is limited to helping in estimating the exterior orientation parameters of the aerial images and the 3D coordinates of tie points. The precision difference arising from the spatial resolution disparity between aerial and satellite imagery is addressed by tuning the stochastic model using variance component estimation. The method was evaluated using a block of aerial images captured by an UltraCamD camera and a WorldView II satellite image, both of which covered a part of Tehran city. The indirectly refined RPCs of WorldView II were utilized, and the required tie points between the aerial and satellite images were measured manually by an expert.

Results and Discussion: The results of this research were evaluated in three different scenarios. The results demonstrated the success of the proposed method in reducing the necessity for GCPs in aerial triangulation and preventing to face of its problems of deficiencies in the definition of the ground coordinate system. Using this method, even with a single GCP, has achieved competitive accuracies compared to using multiple control points in bundle adjustment. However, if there is an appropriate number of control points in the bundle adjustment, this solution will not have a noticeable effect on the aerial triangulation results.

Conclusion: Integrating HRSIs and their refined RPCs, as described, effectively reduces the dependency on GCPs in the aerial triangulation process. Future research should investigate the effect of the spatial resolution ratio of satellite and aerial images on the results, as well as the potential of using stereo satellite images to enhance the capabilities of this method further.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .