| Owing to its low cost and flexibility,the applications of small size Unmanned Aerial Vehicle(UAV)are expanding rapidly and the relevant technology is flourishing.Flying at low altitude,small UAVs manage to achieve high ground resolution,thus making the aerial images excellent supplement to the satellite and high altitude aerial imagery.However,single low altitude aerial image suffers the disadvantage of small field of view,obvious perspective distortion and information redundancy.This thesis focuses on the issue of organizing and displaying low altitude aerial image in big amount.This thesis propose to orthorectify and stitch the low altitude aerial images to form wide view field,scale-unified panoramas of object areas with high ground resolution.In ortho images,the perspective distortion is diminished,the two-dimensional scale information is maintained,the field of view is enlarged,the redundant information is dispelled,and thus visual information of surveyed area is well synthesized and displayed.Research in this thesis is of high value in numerous field such as military intelligence and battle field command,geography information survey,city planning,agricultural and forestry monitoring,etc.To be specific,this thesis contribute in the following aspects.First,this article reviewed and compared common methods regarding ortho-image generation.Camera calibration and reprojection based orthorectification method,Digital Elevation Model(DEM)based ortho-image generation method and new view synthesis based ortho-image generation method are researched and analyzed.Moreover,comparative study of the valid scenarios,merit and demerit of these methods have been conducted.Second,the thesis proposed ortho-image generation technique using multi-view dense matched point cloud.Through three-dimension feature matching and calibration,the point cloud is produced and densified,which is then used to create ortho-images by vertical projection to the ground plane.Third,this article optimized stitching method for ortho-images.Exposure smoothing and color blending were conducted to better the seam.An orthorectified panorama was produced after these steps.Finally,this thesis designed flight plans for the campus scene and acquired low altitude aerial image data using a quadrotor.By conducting grouped orthorectification and stitching,the proposed algorithm is verified. |
