The Research Of Image Matching And Image Correction In UAV Photogrammetry

UAV(Unmanned Aerial Vehicle) becomes a new carry platform because of its quickdevelopment. It plays an important role in all trades. It is used widely in many fields includingmilitary, traffic, disaster monitor, photogrammetry and so on. At present, UAVphotogrammetry technology is one of the research spots at both home and abroad. Imagematching and correction are difficult research points in the field of UAV photogrammetry. Thesolution of image matching and correction is the basis of a series of subjects, such as imagesplicing, large-scale topographical survey and three-dimensional vision. This thesis focusesmainly on image matching and correction.Firstly, the special flying attitude in which images are captured of UAV is analyzed. Dueto features such as small size, light weight and vulnerability to environment, UAV is unstablewhile flying. The flying altitude of UAV is lower than that of MAV. Unmeasurable camera isloaded in UAV. The images which are captured have many characteristics, for example, smallin size, large in quantity, high in resolution, large in angle and so on. Secondly, according tothe trait of images, a classic SIFT matching algorithm which is good robustness at scale,rotation, and lightning is presented in this thesis. The self-correlation of image represents therelationships among pixel grayscale and describes the texture features which are shown fromgrayscale. The thesis proposed a SIFT matching algorithm based on semi-variance functioncombining SIFT with self-correlation. The experimental result shows that the matching rate ofthis algorithm is2%higher than the classic SIFT algorithm; and the matching time reduces23%.Finally, image correction is studied theoretically. The result of image matching pair fromsemi-variance function SIFT algorithm is used to rectify image with epipolar constraint. Thenimages are projected to the same plane. According to internal and external orientationelements, images are converted to geodetic coordinate system using collinear equation toimplement image correction.