Determination Of Geometric Model For IKONOS Satellite Stereo Images And Analysis Of Accuracy In 3D Reconstruction

In recent years, the technology for high resolution remote sensing has been improved greatly. Aerial photography has got rid of the phase that it can only produce small area topographical maps, and has become a way of mapping large area topographical maps. With the successful launch and expansive application of IKONOS satellite imagery, much attention has been paid to the field of positioning with high resolution stereo images. However, satellite photogrammetry can be more complicated than aerial photogrammetry for that the former uses linear sensor arrays that scan a ground strip while the satellite moves forward. The high orbital altitude and the narrow beam of light cause the strong correlation between the geometric parameters. It is too much difficult to solve the orientation parameters. So, a kind of replacement sensor model (i.e. Rational Function Model) comes with the tide of fashion.In this thesis, Rational Function Model (RFM) which is the physical sensor model of IKONOS satellite has been analyzed. The investigation performed in this thesis and relevant conclusions are outlined as follows:First, this thesis analyzed the imaging mechanism of IKONOS satellite and some typical kinds of sensor models. The Rational Function Model which is the sensor model of IKONOS satellite is emphasized.Second, two ways of solving the Rational Polynomial Coefficients which were terrain independent and terrain dependent solutions were investigated. And the three-dimensional determination procedure based on these two methods was examined separately.Third, terrain independent solution based on the physical sensor model and no ground information was involved. The research was studying the different order of RFM by comparing the first-order terms, the second-order terms and the third-order terms. The testing results show that the accuracy of first-order terms was not good, and the second-order terms accuracy has improved a lot, and the third-order terms accuracy keep the same as the second-order terms. So, the second-order RFM did the work strongly.Forth, concerning that the precision of terrain dependent solution has strong connection with the ground information involved, different distributions of GCPs were set to study the solved RPC accuracy and the effect of GCP distribution to the RPC accuracy. The testing results show that the best RPC accuracy can be acquired when using the distribution of GCP in the four corner of the image.Fifth, analyzing the three dimension reconstruction for IKONOS stereo images and evaluating the accuracy. The results show that the mapping precisions based on terrain independent RPC solution were X: 1.199m, Y: 0.866m, Z: 0.842m, and the mapping precision based on terrain dependent RPC solution was X: 0.456m, Y: 1.232m, Z: 1.130m. Therefore, if the physical model or the RPC is unavailable, users can use this method. In addition, in the DTM extracted, the precision in urban area is better than that in mountainous areas. It can be attributed to difficulty in image matching across mountainous areas with vegetation coverage.