Autonomous Navigation Method Research For Pinpoint Landing In Deep Space Exploration

In this paper, the author mainly reseatch the vision aided inertial navigation method, which is a hot research at home and abroad and can improve the accuracy of navigation to meet the require of pinpoint landing.In this paper, the autonomous navigation method involves two aspects: image processing and filtering.Firstly, in the image processing, in order to acquire the corner features of image, the author firstly studyed the Harris corner detection method and test the algorithm's sensitivity with image noise and detection accuracy; secondly researched the transformation between two images, this article assumes that the transformation between the image acquired when the vehicle falls down to the land and the map is a plane-to-plane transformation, this assumption greatly simplifies the transformation algorithm, but does not affect the accuracy of the algorithm; finally the author has researched the image matching method, application phase correlation method and polar coordinate conversion method to estimate the image's translation, rotation and scaling relationship, combined with the matching algorithms in time domain to meet the matching accuracy of 1 ~ 2 pixels.Secondly, in the navigation filter aspect,with the image processing method, we can determine the global coordinates of the feature points in images taken at the landing phase. Combination these coordinates with inertial measurement, we can use the Eextended Kalman Filter to get the more precisioner navigation information than inertial which can inhibit the accumulation of inertial navigation error and drift, improving the inertial navigation accuracy. Finally, in order to analyze the performance of the navigation algorithm, the author has used mathematical simulation to acquire the relationship between the accuracy of the method and the image processing frequency, the number of features, the map's resolution and the mismatch.