Abstract:
Generally, there are two approaches applied with high resolution images in order to correct errors and distortions. The first one is Parametric Model, which essentially requires sensor alignment parameters, and sensor property parameters. Unfortunately, these parameters are non-disclosure; therefore, it is quite difficult to acquire all those data. The other approach in Non-Parametric Model. This method focuses on applying simple mathematical equations to determine correlations between image planes and object plane on Earth disregarding with the sensor alignment. This research aims to assess the high resolution satellite image horizontal accuracy on standard product of QuickBird. This type of image contains positioning accuracy of 14 meters (RMSE) with image resolution of 0.60 meters. The study focuses on a specific area in Bangkok, applies non-parametric model with 1st, 2nd, 3rd order polynomial formations, accordance with ground control point (GCP) obtained from GPS survey. This also includes comparing satellite images with Bangkok Map 1:4000. The result illustrates that introducing the adjustment with 2nd order polynomial equations with 10 ground control points provides the best horizontal accuracy. The test obtains the degree of horizontal accuracy at 1 meter, and 1.60 meters under the condition that appropriate GCPs from GPS survey and Bangkok map scale 1:4000 are applied, and the texture and details in the image are clear and complete comparing with Bangkok map scale 1:4000. According to the outcome from the research which indicates high horizontal accuracy, we may conclude that high resolution satellite images like QuickBird contain appropriate potentials for large scale map revision in small specific areas.