The Uavs Image Automatic Registration And Splicing Method

UAV (Unmanned Aerial Vehicle) remote sensing techology has some advantages----mobility, timeliness and so on. So it can remedy the detects of satellite remote sensing and aerial remote sensing on timeliness. It can provide the information for disaster and emergency command in the first time. Nowadays, UAV remote sensing has become the focus of attention. Typically, the single remote sensing image can not cover the whole area being researched and is difficult to get the global information, which requires the registration and mosaic of multiple images. Therefore, with the mass low-altitude remote sensing data, how to finish quick registration and automatic mosaic is the current problem needed to be resolved.Based on UAV images of non-geographic coordinate and non-metric camera data, this paper researched feature point detection, image registration and mosaic and other related theories and methods, quickly finished the image automatic registration and mosaic of the whole strip. Firstly, the author made feature detection with SIFT operator and made additional feature detection with Harris operator,which is to non-obvious-feature areas (smaller land, larger water).Then, the author filtered these detected feature points, made rough matching, researched the RANSAC algorithm deeply and removed a large number of error matching points with the method, which is combined with the RANSAC algorithm and the epipolar constraint. In this paper, chose some different kinds of transformation models to match the two images, such as the affine transformation, quadratic polynomial, Delaunay triangulation and homography matrix. At the same time, compared and analyzed the the effects of these transformation model, and finished the registration and mosaic of the whole strip with the affine transformation.Finally, the author made further prospect to how to finish automatic registration and mosaic of large-scale UAV images quickly with the POS data.