RFM-Based Image Matching Technique Using Object-Space Approach

Abstract

The Rational Function Coefficients (RFCs) provided with the high-resolution satellite imagery show an accuracythat is relatively low and inconsistent with their original image quality. Also, there have only been a few detailed researches about 3-D terrain data generation using Rational Function Model (RFM) . In this study, we suggest an improved matching algorithm which uses RFM as an image to object transformation model and introduce the detailed procedure of generating 3-D surface models from the RFM for non-experts in photogram-metry and remote sensing. The proposed matching scheme constructs the Piecewise Matching Line (PML) by using elevation range in the object space. The constructed PML determines the initial location and confines the search space in image matching. The proposed scheme is applied to SPOT-3 stereo images which supply the detailed satellite orbital information. The RFCs of SPOT-3 are obtained using the terrain-independent approach. The 3-D positional accuracy of the calculated RFCs is about RMSE of 2 meter compared with that of the physical sensor model. Finally, DEM over test area is generated using the proposed matching scheme. DEMs generated from the physical sensor model and from 3 arc-second Digital Terrain Elevation Data (DTED) are used to evaluate the relative and absolute accuracies of the generated DEM. The result shows RMSE of 7.51 meter in relative accuracy and 16.50 meter in absolute accuracy.