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

Abstract

data, different research studies have addressed the potential of Thirteen panchromatic (Pan) and multiband (XS) Ikonos Geo- high-resolution imagery for mapping. A research study at the product images over seven study sites with various environ- National Mapping Agency of Great Britain used simulated 0.2ments and terrain were tested using different cartographic data m and 1-m Pan images derived from 1:7,500-scale aerial photos and accuracies with a 3D parametric model developed at the and XS 4-m images from the Compact Airborne Spectrographic Canada Centre for Remote Sensing, Natural Resources Canada. Images (Ridley et al., 1997). Some results indicated that the The objectives of this study were to define the relationship future high-resolution satellite imagery would have potential between the final accuracy and the number and accuracy of for improving the existing National Topographic Database of input data, to track error propagation during the full geometric Great Britain from 1-m panchromatic images as well as autocorrection process (bundle adjustment and ortho-rectification), mating the detection of topographic feature change from 4-m XS and to advise on the applicability of the model in opera- images. In another research study, a theoretical analysis based tional environments. on in-track and across-track stereo mapping techniques demonWhen ground control points (GCPs) have an accuracy strated that high-resolution satellite imagery could be used for poorer than 3 m, 20 GCPs over the entire image are a good the generation and updating of national mapping products in compromise to obtain a 3- to 4-m accuracy in the bundle the United States, but only if photogrammetric processing were adjustment. When GCP accuracy is better than 1 m, ten GCPs employed (Li, 1998). Based on this theoretical analysis, an are enough to decrease the bundle adjustment error of either evaluation of the potential accuracy of ground points was panchromatic or multiband images to 2 to 3 m. Because GCP performed using Ikonos stereo images simulated from aerial residuals reflect the input data errors (map and/or plotting), photos (Zhou and Li, 2000). Without ground control points these errors did not propagate through the 3D parametric (GCPs), a positioning accuracy of 12 m in the three axes (X, Y, model, and the internal accuracy of the geometric model is and Z ) was achieved, and with 24 GCPs the accuracy was thus better (around a pixel or less). improved to 3mi n planimetry (X and Y ) and 2 m in elevation. Quantitative and qualitative evaluations of ortho-images More recently, preliminary accuracy tests with real images were thus performed with either independent check points or having a viewing angle of 39 and using a3D parametric geooverlaid digital vector files. Generally, the measured errors metric correction model were performed with few precise confirmed the predicted errors or were even slightly better, GCPs (Toutin and Cheng, 2000; Davies and Wang, 2001). The and a 2- to 4-m positioning accuracy was achieved for the results of the two research studies demonstrated positioning ortho-images depending upon the elevation accuracy (DEM accuracies of 2 to 3 m and confirmed the high mapping potenand grid spacing). To achieve a better final positioning tial of Ikonos images. accuracy, such as 1 m, a DEM with an accuracy of 1 to 2 m This paper expands on the above-mentioned preliminary and with a fine grid spacing is required, in addition to well- results related to the geometric processing and ortho-rectificadefined GCPs with an accuracy of 1 m.