Elevation Information Extraction And Accuracy Ananysis Based On INSAR

In the last two decades, Interferometrc Synthetic Apenture Radar (INSAR)technique developed as an advanced space-based observations technique whichperforms well in topographic mapping. Recently, INSAR received moreattentions due to its ability to obtain DEM in all-time.The application of INSARtechnique is influenced by development of techniques in interference filteringand phase unwrapping and data qulity.This dissertation studies the spaceborne INSAR techniques, due to the lackof accurate DEM data of the observation area was unable to check the accuracyof elevation information extracted, we made a more systematic study of INSARfor a number of stages in the DEM process.The research work is mainlyconcentrated in the image registration and interference filter step algorithms andparameters of the final height accuracy, based on the coherence and residual toachieve qualitative and quantitative analysis of the results, improve lowcoherence method precision and mismatch.The detailed contents are as follows:First, we depth study of the coherence method commonly used in imageregistration steps and maximum spectral method. Experimental results showedthat the maximum spectral method compared to the coherence method accuracyis slightly higher, but not obvious in the case of noisy interference fringes andcoherence method is prone to the phenomenon of mismatch, then this paper, thecoherence method image offset calculation error, the matching accuracy is nothigh to improve the coherence method to achieve improved accuracy, toeliminate the mismatch. Finally, based on the orbital parameters using theimproved coherence with the quasi-experimental validation, coherence of theevaluation criteria to prove a substantial increase in registration accuracy.Secondly, we study of the mean filtering, median filtering, cycle round themean and median filtering, maximum likelihood filering and other interferogramfiltering algorithm. Residual points for the evaluation of the filtering effect of thestandard and compare the different interference filter algorithm on the elevationaccuracy. by the experimental comparison, circular periodic mean, median filtering and maximum likelihood filter can better retain the edge information ina window-based filtering algorithm, maximum likelihood filtering effect is thebest, the residual points is relatively small.Finally, we use Doris open source software written by team of theUniversity of Delft, The Netherlands to processing real SAR data step-by-step,and achieve the extraction of elevation information. We Summarize the mainfactors affecting the height accuracy experiment, including image resolutionparameters calculation accuracy, the registration accuracy, the filtering effect andgeometric caused by distortion topography characteristics, radar stack cover,shadows.