Semi Global Dense Matching Based On Hough Transform And Fourier Transform

In dense matching,binocular stereo vision is one of the important directions of today's development.Binocular camera can get the depth information of additional space points through disparity map,and obtain the coordinates of space points in camera coordinate system.In binocular stereo vision,semi global matching algorithm is widely used.Compared with local matching algoritm,it greatly improves the accuracy and efficiency.In this paper,Hough transform and Fourier transform are introduced into the semi global dense matching to process the UAV aerial images,which can improve the accuracy of linear object recognition,weaken the salt and pepper noise and reduce the holes and deformation in the point cloud,so as to improve the quality of the dense point cloud,and improve the accuracy and efficiency of the semi global matching.In this paper,two groups of experimental data are used to test the accuracy and efficiency of the improved data processing method on OpenCV platform(1)Hough transform is used to improve the recognition accuracy of linear objects.Through a series of linear transformations,Hough transform transforms the pixels in the image from a space rectangular coordinate system through rotation and translation,and projects them to the polar coordinate system.It iuaps the lines or curves with the same characteristics in the original space to other spaces to form sinusoidal curves,and transforms the peak detection from a geometric problem to a statistical problem.The gray images of the left and right images of binocular vision are detected by Hough transform.(2)After the data processing of Hough transform,Fourier transform is used to transform the line detection image obtained by Hough transform into Fourier transform and inverse Fourier transform,so as to weaken the influence of salt and pepper noise,and improve the quality of dense point cloud and improve the efficiency,the efficiency is increased by 0.299%.(3)After Hough transform line detection and Fourier transform processing,the semi global dense matching algorithm is used for image matching.The four necessary steps of dense matching are performed to obtain the left and right disparity map.The disparity calculation is based on the left view to obtain the left disparity map.Then,dense point clouds are obtained from the disparity map formed by consistency detection,median filtering and semi global dense matching algorithm.The noise and execution efficiency of the two methods are compared.(4)The experimental results show that the linear characteristics of semi global dense matching point cloud are significantly improved after processing UAV aerial photography images by Hough transform and Fourier transform,and the quality of point cloud is also significantly improved.This paper has 46 figures,4 tables and 64 references.