Trajectory modeling for satellite image triangulation

The model components of satellite photogrammetry consist of a trajectory model, projection equations and parameter subset selection. The trajectory model is important because subsequent estimation is performed by making corrections or refinements to this initial path. However, satellite imagery products are provided with diverse formats of support data having different types, representations, frequencies and conventions. Among the three components of the sensor model, the construction of the position and attitude trajectory is closely linked with the availability and type of support data. In order to build a physical sensor model compatible with the metadata, a number of trajectory models have to be set up, and the influence of each trajectory model has to be analyzed. In order to investigate these issues in a practical way that is tied to real data, we show how trajectory models can be implemented based on support data from six satellite image types: Quickbird, Hyperion, SPOT-3, ASTER, PRISM, and EROS-A. Triangulation for each image is implemented to investigate the feasibility and suitability of the different trajectory models. Also, to evaluate the result, we used the leave-one-out cross-validation(LOOCV) method that enables effective use of a group of point observations, and provides independence to the check point selection and distribution. The results show the effectiveness of some of the simple models while indicating that careful use of dense ephemeris information is necessary. These results are based on having a number of high quality ground control points.