Research On High Precision Landmark Matching Technology Of Satellite Image Geometric Positioning

Landmark matching refers to calculate the offset of the landmark from the satellite image,which is matched against the landmark template image,then extracting the location of the landmark from the template image in the satellite image.The classic landmark matching algorithms,such as the landmark matching algorithm based on the cross correlation and the landmark matching algorithm based on gradient energy,cannot perform well when the satellite image rotation occurs.Focus on the problem of the landmark matching when the satellite image rotates,a landmark matching algorithm based on edge information is proposed,the special research contents are as follows:(1)Introducing the image filtering technology based on partial differential equations.The simulation results show that regularization P-M diffusion method and tangent direction diffusion method with edge stopping are capable of suppressing noise in the image,while can maintain edge information of the image clarity.In addition,the technology,such as edge detection of the grey image technology,landmark matching techniques based on cross correlation and based on gradient energy and template matching algorithm based on the generalized Hough transform(GHT),are introduced.(2)Proposing a landmark extraction algorithm based on the connected domain analysis of edge points.On the one hand,Canny algorithm is improved by using regularization P-M diffusion method instead of the Gaussian filter method and the dual threshold method based on histogram cancavity analysis instead of a fixed threshold.On the other hand,filtering the edge image according of the number of edge pixels from eight-connected domain,which suppresses further spurious edges from the edge image.The landmark extraction algorithm is proposed based on the above both aspects.The simulation results show that the algorithm can effectively extract the landmark information from the satellite image while suppressing spurious edges.(3)Proposing the GHT algorithm based on the U-chord curvature.Aimed at two inadequate respects of existing GHT algorithm: defining point spread function(PSF)rather than building R-table;geometry calculations using gradient angle.Therefore,a GHT algorithm with anti-rotation properties is proposed using the U-chord curvature and the slope direction.The simulation results show that the algorithm has better anti-rotation and anti-noise.(4)Proposing a landmark matching algorithm based on GHT.The proposed GHT algorithm is applied for landmark matching.The simulation results show that the landmark from the satellite image relative to the landmark from the landmark template image rotates whether or not,the algorithm both has better matching performance.