Geometry Processing For The UAVRS Image Based On Dual Quaternion

In recent years, due to its flexibility, low cost, high resolution, high risk area detection and so on,UAV remote sensing systems have been widely applied to military reconnaissance, natural disastermonitoring and assessment, land use investigation and other national defense applications. The keyproblem of UAV image mosaic is to define the mathematical relationship between the UAV imageplane coordinate and ground coordinate system, so as to extract3D information from UAV images.However, UAV image has the disadvantage of instable attitude, small frame and large quantity etc, sothe traditional method of aerial photography are often unable to proceed. How to solve this problem isa current research hotspot of the UAV remote sensing image processing, and it is also the mainresearch direction of this paper.This paper proposes a novel UAV image geometry processing method using dual quaternion todescribe the position and attitude of UAV images. A dependent relative orientation model based ondual quaternion is established to stitch UVA images without control points. Then establish theGPS-supported bundle adjustment model based on dual quaternion, combine GPS data andobservations to slove the transformation between image points and ground points. Experimentalresults verify the correctness of the proposed method. This research will be of great value to UAVimage mosaic and accurate target positioning based on UAV images.