Fast Automatic Stitching for Images of Unmanned Aerial Vehicle in Emergency Response

With the demand of emergency response,in order to obtain full cover image of remote sensing area fast,automatically and stably,this paper proposes a suitable stitching method of image sequences from unmanned aerial vehicle(UAV).A position and orientation system with low accuracy and a non-metric camera without calibration mounted on UAV can photograph the interested area to attain the image sequences and fast automatic registration.With the global ground points and the corresponding projection points,the whole optimum error equations are calculated using Levenberg-Marquardt algorithm for nonlinear iteration of the Homography matrices and unrestrained projective centers.Then,the parallel geometric seeking algorithm of stitching line is applied;image interpolation and image fusion are processed for whole automatic image outputting.Through the disaster emergency response experiment based on UAV remote sensing,the 0.05m resolution full cover image with less than 3 pixel of positioning accuracy are obtained from 450 images in an hour.The fast automatic algorithm proposed in this paper is appropriate for image sequences from UAV,which can provide effective technology support for emergency response.