Geometric Accuracy Assessment of QuickBird Basic Imagery Using Different Operational Approaches

Abstract

New very high-resolution space satellite images, such as QuickBird and Ikonos, open new possibilities in cartographic applications. This work has as its main aim assessment of a methodology to achieve the best possible geometric accuracy in orthorectified imagery products obtained from QuickBird basic imagery which will include an assessment of the methodology’s reliability. Root Mean Square Error (RMSE), mean error or bias, and maximum error in 79 independent check points are computed and utilized as accuracy indicators. Ancillary data was generated by high accuracy methods: 1) check and control points were measured with a differential global positioning system, and 2) a dense digital elevation model (DEM) with grid spacing of 2 m and RMSE-sub-z of about 0.31 m generated from a photogrammetric aerial flight at an approximate scale of 1:5000 that was used for image orthorectification. Two other DEMs with a grid spacing of 5 m and 20 m were also used. Four 3D geometric correction models were used to correct the satellite data: 2 terrain-independent rational function models refined by the user, a terrain-dependent model, and a rigorous physical model. The number and distribution of the ground control points (GCPs) used for the sensor orientation were studied as well, testing from 9 to 45 GCPs. The best results obtained about the geometric accuracy of the orthorectified images (2-D RMSE of about 0.74 m) were computed when the dense DEM was used with the 3-D physical and terrain-dependent models. The use of more than 18 GCPs does not improve the results when those GCPs are extracted by stratified random sampling.