Error Tracking in Ikonos Geometric Processing Using a 3D Parametric Model

<p><i>Thirteen panchromatic (Pan) and multiband (XS) Ikonos Geo-product images over seven study sites with various environ-ments and terrain were tested using different cartographic data and accuracies with a 3D parametric model developed at the Canada Centre for Remote Sensing, Natural Resources Canada. The objectives of this study were to define the relationship between the final accuracy and the number and accuracy of input data, to track error propagation during the full geometric correction process (bundle adjustment and ortho-rectification), and to advise on the applicability of the model in operational environments.</i></p><p><i>When ground control points (GCPs) have an accuracy poorer than 3 m, 20 GCPs over the entire image are a good compromise to obtain a 3- to 4-m accuracy in the bundle adjustment. When GCP accuracy is better than 1 m, ten GCPs are enough to decrease the bundle adjustment error of either panchromatic or multiband images to 2 to 3 m. Because GCP residuals reflect the input data errors (map and/or plotting), these errors did not propagate through the 3D parametric model, and the internal accuracy of the geometric model is thus better (around a pixel or less).</i></p><p><i>Quantitative and qualitative evaluations of ortho-images were thus performed with either independent check points or overlaid digital vector files. Generally, the measured errors the predicted errors or were even slightly better, and a 2- to 4-m positioning accuracy was achieved for the ortho-images depending upon the elevation accuracy (DEM and grid spacing). To achieve a better final positioning accuracy, such as 1 m, a DEM with an accuracy of 1 to 2 m and with a fine grid spacing is required, in addition to well-defined GCPs with an accuracy of 1 m.</i></p>