Research On The Key Technology For Scene Matching Aided Navigation System Based On Image Features

The key technology for scene matching aided navigation system based on image features was studied in this dissertation. Moreover, the theories, algorithms and ideas were verified through simulation flatform and experiments.The demands of image matching algorithm for scene matching aided navigation system was analysed in this dissertation, that is precision, real-time and robust. The weighted Hausdorff distance image matching algorithm which is suited to scene matching aided navigation system was proposed, and also the weighted computation fomula was deduced. At the same time, to satify the requirement of real-time and robust of navigation system, aimed to the problem of feature point data being redundant, the pre-proccessing method of extracting bifurcations from edage feature binary image was put forward. It was combined with weighted Hausdorff distance algorithm, and a set of basic image matching theory and algorithm was completed.For the problem of global searching affecting speed of matching, the genetic algorithm was utilized to speed the searching process of image matching. Combined effecitively with weighted Hausdorff distance image matching algorithm, it can quickly majorized search the best matching point. Furthermore, considering the weighted Hausdorff distance image matching algorithm can only locate in the range of pixel, the least square algorithm based on bifurcation was used to more improve the precision of system. So, the theory and detail realized idea of a multi-level matching algorithm were systematic proposed for scene matching aided navigation system. This algorithm can integrally improve the accuracy, real-time and robust of the scene matching adied navigation system.When the scene matching aided navigation system is in the stage of searching, it has a rather large initial error. Because there are much geometric distortion between real-time image and reference image, the matching algorithm proposed above can finite correct it. So, from the idea of space transforming, the log polar transform was utilized. With the characteristic of error deviation of inertial navigation system, based on image feature extraction, the theory and detail performing idea of a center point with its 4 neighbours log polar transform image matching algorithm were proposed. This algorithm can effectively tolerate rotation and scale effection to the precision of the image matching in the initial stage. It also gives a reference to the next weighted Hausdorff distance image matching algorithm. For broadening the applied range of scene matching aided navigation system, a performing program of SAR/INS/TERCOM integration navigation system was proposed. The problem of image matching not giving heigh measuring information was also solved. Furthermore, aimed to actual image matching information output and tercom output having different interval and delay respectively, a cooresponding multi-information fusion filtering algorithm was proposed. It can satify the actual applied demand of SAR/INS/TERCOM integration navigation system and give engineering solving idea for the problem of filtering for image matching applying to integration navigation system.An integration navigation system simulation flatform using scene matching aided navigation was built. And based on engineering actual situation, the realizing of scene matching aided navigation simulation system and verifying image matching algorithm were studied. The main structure and functional modules of simulation flatform were put forward. And the image matching algorithm and filtering algorithm proposed in this dissertation were embedded in the simulation flatform. The simulations show that the algorithms proposed in this dissertation are effective. These works provide the theory guide and technology support to algorithm transforming engineering principle model machine. It has a rich value of engineering applying.