Research On The Real-time And Accurate Scene Matching Algorithm Based On Local Invariant Features

The scene matching algorithm in the inertial integrated navigation system based on image local features is studied in this dissertation. Moreover, the theories, algorithms and ideas are verified through simulation platform and experiments.It is important that image matching algorithm should be satisfy the demands of scene matching aided navigation system, that is precision, real-time and robust, and also should be robust to occlusion, background clutter and noise, invariant to various image transformations due to translation, rotation, scale, affine deformation, difference in illumination, object movement, and change in viewpoint. In consideration of the part Hausdorff distance as the similarity measuring to images match and localize, matching accuracy is limited, the speed is relatively slow, ability of rotation and noise resistance are poor, and difficult for getting the precision rotation transform parameters between images. So the local invariant feature is applied to scene matching aided navigation system.SIFT(Scale Invariant Feature Transform) features are highly distinctive and invariant to various image transformations due to rotation, scale, and change in viewpoint. In the paper, we propose the scene matching algorithm based on SIFT features for inertial integrated navigation. We using SIFT features and feature matching algorithm, gives a reliable and fast image matching algorithm. This algorithm can improve the accuracy, real-time, and stability of the scene matching aided navigation system.For the scene matching problem of SAR images by the geometric distortion and under the influence of serious speckle noise, the SURF(Speeded-Up Robust Features) algorithm of scene matching aided navigation is proposed. SURF approximates or even outperforms previously proposed schemes with respect to repeatability, distinctiveness, and robustness, yet can be computed and compared much faster. The experiment result demonstrates that the SURF in the SAR/INS integrated navigation system with good application.At the same time, in the consideration of SIFT and SURF algorithms have certain errors, obtained the feature point position existence distortion change, the RANSAC algorithm is used to remove the false and low precision matching points, and the least square algorithm based on image points extraction is applied for getting both the aircraft position errors and course deviation.Finally, based on engineering actual situation, using scene matching aided navigation system simulation platform, and the image matching algorithm proposed in this paper for the source of information are studied to amend the error. The simulation shows that the algorithm proposed in this dissertation is effective.