The Research On Automatic Relative Orientation And Model Connection Of The Unmanned Aerial Vehicle Remote Sensing Images With High Precision And High Reliability

In recent years, as an effective complement to the satellite remote sensing and conventional aerial photogrammetry, the Unmanned Aerial Vehicle (UAV) has been widely used in geological environment and disaster investigation, dynamic monitoring of land use, map updating and other fields. During the period of disaster relief in 5.12 Wenchuan earthquake, Yushu earthquake and the powerful mudslides happened in Zhouqu Country, Yingjiang earthquake, the UAV with its many advantages acquired the high resolution remote sensing data of the disaster area for the first time. However, the captured images have problems such as small format images, excessive quantities, over tilt angle, irregular forward and side overlap etc. The exist of these problems brings a series of trouble into image matching, aerotriangulation and some other relative subject. To solve the problems existed in the UAV remote sensing data, this paper has researched mainly on the following work:1. Study how to correct the original remote sensing images based on the distortion model. Introduce the methods of aerotriangulation. Analyze the characteristics of the POS data and present an appropriate aerotriangulation way for the UAV remote sensing images.2. Describe the intensity correlation matching and the least square algorithm. Research the principles and processes of SIFT(Scale Invariant Feature Transform) deeply. The SIFT algorithm is applied to the automatic image matching in order to improve the success rate.3. Analyze the reason for the failure of automatic relative orientation in the previous photogrammetric software. Put forward an appropriate strategy of automatic relative orientation for the UAV. Discuss the solution to eliminate the gross error robustly and ensure the point position evenly distributed.4. Analyze the reason for the position of model connection points unevenly distributed and model connection failure, propose the mean square error of model connection is should be relaxed, research how to eliminate the false model connection points and make the points position evenly distributed.Image distortion correction, automatic relative orientation and model connection of the UAV images in this research have been used in the integrated system for high resolution remote sensing image processing—PixelGrid, and they have been widely applied in the production units with highly practical served as a part of three-dimensional triangulation belonging to the UAV remote sensing image processing module of PixelGrid.