Matching Suitability Of SAR Imaging Areaes For INS/SAR Integrated Navigation

In SAR scenery matching aided INS/SAR integrated navigation system, matching precision is guaranteed by two key factors, which are SAR image matching algorithm with high reliability and suitable matching areas with high success probabilities respectively. But compared with multifarious SAR matching algorithms proposed, there are very few research works concentrating on the selection of good SAR matching-suitable areas at present. SAR image is radar signal in essence, so besides the similar problems to the selection of matching-suitable areas for optical images, SAR imaging conditions and characteristic constraints of flight platform in INS/SAR should also be considered when choosing good SAR matching areas. Under the background of the analysis above, the core problem of SAR matching area selection for INS/SAR integrated navigation - matching suitability of SAR imaging areas - was studied detailedly in this dissertation.In the first place, as a synthetical analysis to SAR matching suitability problem, a general research framework was constructed combined with SAR imaging process and characteristics of related sensors in INS/SAR integrated navigation system. Under the research framework, basic suitability evaluation indexes for SAR and several influence factors were discussed, and the whole matching suitability problem was separated into three parts, which respectively are how to define reliable range of a matching area, how to construct image suitability measurement function based on a synthesized matching-suitability feature, and how to select matching areas under the influence of distortion by terrain.Afterwards, on the basis of the synthetical analysis, three key technologies were studied emphatically, which were estimation of survey error of SAR ground observing position, construction of synthesized matching-suitability feature, and selection of matching areas under image distortion by terrain. Concrete contents and contributions are as follows.(1) An estimate method of survey error of SAR ground observing position was proposed, which can serve for the determination of reliable range of a matching area.Survey error estimation of SAR ground observing position determines whether a reference image area fits matching from the requirement that an observing image should be covered within the reference image area completely. Firstly, based on radar imaging principle in the range direction, a theoretical model for transfer relation between ideal position signals from the centre of an aircraft to its corresponding ground observing point was established. Then inspired by the fact that all of the most dominant factors to deviation show approximate linear influence characters, a simplified linear model for survey error based on main factors was set up. Furthermore, a computing model of deviation distribution on the basis of the simplified linear model was constructed. Simulating biased estimate of ground survey error between the simplified linear model and the precise theoretical model is less than 10? 8 radian(equal to 6mm), estimation result of MSE(mean square error) of ground survey deviation between computing value through the deviation distribution computing model directly and Monte-Carlo statistic value under 8000 samples through the precise theoretical model is less than 10?7 radian(equal to 6cm), and therefore deviation estimation with high accuracy in zero-sample circumstance is achieved when giving MSE of dominanting factors.(2) A construction algorithm of synthesized matching-suitability feature was established based on SAR image gray information.A synthesized feature construction algorithm was established based on SAR image gray information combined with advantages of expert experiences and computer fast searching. On the one hand, several primary suitability features were designed according to characteristics of image textures, SAR imaging and SAR matching algorithm, and this course reflects the expert experiences; on the other hand, a flexible synthesized feature was constructed on the basis of synthesized operation expression composed of primary suitability features, and afterwards, optimization searching based on genetic algorithm was introduced to find synthesized suitability feature with the highest fitness. As a result, expression space is expanded greatly when elevating the searching effectiveness. Simulated experiment results based on real C-band and P-band AIRSAR data show that the synthesized feature gained can reflect the matching suitability of SAR gray images accurately, and matching probabilities of selected matching areas reach to 99 %±0 . 5%.(3) A synthetical good matching area selection method under image geometric distortion was proposed.A synthetical matching-area selection method was established combined with constraint from terrain wave. Firstly, it was tested from theoretical analysis and experimental simulation that observing image distortion influences matching suitability of a certain area more seriously than reference image distortion does in the practical application of INS/SAR integrated navigation. Then based on SAR imaging principle, an observing image calibration algorithm was designed when DEM of the observing image area can not be obtained precisely, which only requires the distribution character of terrain within the corresponding reference image area. Finally, change relation between success probabilities and values of different terrain features were compared combined with a distribution model of natural terrain, and on the basis of which, an integrated synthetical matching-area selection process was proposed. Simultaneously, two conclusions were gained, which respectively were (i) standard deviation of a terrain influents matching suitability of a SAR image area from the angle of geometric distortion more effectively compared with a terrain fractal parameter; (ii) low standard deviation of a terrain is a necessary condition to acquire high success probability.Ultimately, based on the theoretical research, simulated environment was constructed under which validity of the algorithms was tested through an integrated matching-area selection experiment. During the process of the experiment, real AIRSAR data were used and related parameters from vehicle platform and imaging sensors were set according to expert experience. Simulated results under unified mutual-information based matching algorithm show that entrance probabilities of selected matching areas are over 96%, success probabilities once entered even reach to 100%, and success probabilities on the consideration of geometric distortion are still up to 90%, which sufficiently validates the rationality of the established research framework and the effectiveness of the proposed algorithms.