Accurate Geometric Positioning Of Spaceborne SAR Remote Sensing Imagery

Since the technology of synthetic aperture radar (SAR) was open in 1960s, its role in Earth observation has been revealed gradually. In recent years, with the rapid development of Earth observation techniques, high-resolution, multi-sensor, multi-polarization, multi-band spaceborne SAR remote sensing data emerging continually, the traditional photogrammetric study object is not limited to optical data source any more. Because of the unparalleled observation ability of all-weather, all-time, large-scale, strong penetration, Synthetic aperture radar has been widely used in the field of land resource survey, change detection, urban planning and geographic information services and so on. Meanwhile, spaceborne SAR imagery, as a useful complement to the optical imagery and making the real-time natural disasters data processing becoming possible, will play a more important role in the future. Image geometry positioning technology is the basis of various remote sensing applications, so in order to fully exploit the potential of SAR imagery, it must firstly establish the accurate geometric relationship between image coordinates and the corresponding ground coordinates. Therefore, aiming to the special characteristics of SAR imagery, combined with the principle of radar photogrammetry, it is of great importance and meaning to carry out the research on the accurate geometric positioning technology for SAR remote sensing imagery.This paper takes spaceborne SAR remote sensing imagery as the main research object, carries out the study of single image positioning with lack ground control points and multi-source SAR imagery bundle adjustment, makes a deep research on the fessibility of the rational function model to alternate the rigorous geometric model, and also performs relevant research on multi-source SAR images matching. Through performing a large number of experiments on spaceborne SAR remote sensing imagery covering many areas of China, the correctness of various algorithm and the fessiblity of the propsed scheme are validated. The main research work and innovative contributions are as follows:1) The positioning error of spaceborne SAR imagery is analyzed with the simulation method. From the results, it can go to the conclusion that compared to the error in satellite position, the error in satellite velocity has greater effect on geolocation. But due to the high accuracy of the velocity in reality, the error in satellite position is the main factor of positioning error practically, which is strongly systematic in the object space. In addition, the total effect reflected in image space of various errors, also has a strong systematic characteristics. Using only a small amount of ground control points and making compensation in image space, can significantly improve the positioning accuracy.2) Considering the difficulty to obtain the ground control points, the single geometric positioning method based on strict geometric model with lack of ground control points is proposed in this paper. It has shown that through the "virtual observation" method, making local optimization to the main orientation parameters of SAR imagery, can improve its positioning accuracy substantially.3) For the malti-sensor SAR imagery automatic matching, the method based on SIFT operator and CRA similarity measure is introduced. The results of the experiments have shown that our algorithm uses the SIFT operator to estimate the rotation angle and the resolution difference, which are then used to make compensation to the matching window image, takes CRA as the similarity measure, incorporates the coarse-to-fine pyramid image matching strategy, and applies quadratic polynomial model supported by RANSAC, this can effectively overcome the large rotation angle between images, significant differences of gray-scale and high false matching rate.etc, to achieve better matching results, which enable the automatic tie point acquisition for multi-sensor SAR image bundle adjustment.4) Propose the RFM model based bundle adjustment method for multi-source SAR imagery. Make a comparison to the bundle adjustment method based on the rigorous geometric model, and at the same time explore the impacts of different conditions of control point’s distribution on them. Experimental results have shown that the method based on the rigorous geometric model has a strong dependence on the number and distribution of ground control points. When distributing dense points on the border rather than sparse points, it can enhance the location accuracy evidently. But distributing more ground points in the middle has little effect on the results. The method based on the RFM model is less dependent on the number and distribution of ground control points relatively. When there are not enough control points and distribution condition is not good, it can achieve better bundle adjustment results than that of the method based on the rigorous geometric model. If enough control points are available and they have a good distribution, the method based on the rigorous geometric model is still the best choice for the bundle adjustment of SAR imagery. The bundle adjustment method based on RFM model, as the complement for multi-source SAR imagery combined geolocation, is also useful.