Research On The UAV Remote Sensing Image Mosaic

The application of unmanned aerial vehicle remote sensing is more and more widelywith its’ high spatial resolution, high efficiency, low cost and high security. With the limit ofthe flight height and high spatial resolution, the coverage of a UAV image is very small, andwe can’t get enough useful information for the whole research region by one image. Thereforewe need to mosaic together all the images.The key technology involved in the process of image mosaic mainly include iamgeregistration and image fusion.Image registration basically determines the success of imagemosaic.Based on the study of relative knowledge,we find feature-based registration methodjust take advantage of the significant characteristics of the image,greatly compress the imageinformation,and thus reduces the computation complexity.In addition,it has a good robustnessto rotation,zooming,noise and so on. In this paper, the author focuses on image registrationbased on feature.This article describes related theoretical foundation and key technologies of imagemosaic.First,the classic feature point detection algorithm presented by previous is summarize.In addition, its advantages and disadvantages are contrasted. In terms of feature extraction:theauthor detection image feature points using SURF algorithm,and make coarse matching withneighboring search strategy.In terms of image registration:we remove the repeated featurepoints the feature point’s pixel coordinates,and then,we exclude part of some mismatchingpoints by slope constraint.Finally,based on perspective transformation,we use RANSACalgorithm to get rid of outer points.By these steps,we ensure the accuracy of imageregistration. For image mosaic and image fusion:we chose homography matrix to matchimages. And then,multi-resolution spline technology is used to achieve seamless mosaicbased on the best stitching line.The paper uses C++simulation experiments to verify matching performance of SURFalgorithm.The results show that SURF algorithm not only maintains good robustness andimmunity,but also has a faster matching speed than SIFT.