Geometric Modelling and Processing of Quickbird Stereo Imagery

This paper discusses geometric modelling and photogrammetric processing methods for high-precision mapping using stereo QuickBird images, and reports research results of data processing in an area in Tampa Bay, Florida,. First, a set of panchromatic “Basic” QuickBird stereo images are evaluated using GPS control points. Systematic errors in the vendor-provided RPCs (Rational Polynomial Coefficients) are found to be 8.8m, 8.7m and 12.6m in the X, Y and Z directions, respectively. Next, a study of the improvement of 3-D geopositioning accuracy is conducted in which a comparison is made between different adjustment models that represent different numbers and configurations of ground control points. The comparison results indicate that a simple adjustment model with only a couple of ground control points is effective for the elimination of systematic errors and for the improvement of 3-D geopositioning accuracy to a 0.7-1m (1-1.5 pixel) level. In addition, a sub-pixel accuracy of 3-D measurement can be achieved for feature points in images that can be sharply identified and, thus, accurately measured. Based on this improved geometric model, a 3-D shoreline was produced semi-automatically in the area.