Research On Mosaic Technology Based On UAV Remote Sensing Image

Image mosaic is a technology to solve the problem of the inadequacy of single camera’s photo. It has brought a lot of help to people’s daily lives and scientific research. With the development of digital image processing and remote sensing technologies, wide field mosaic images have been widely used in computer vision,aerospace application, mapping and patrol pipeline, medical images analysis, remote sensing image processing and other fields. The wide field mosaic images became a hot research spot of image processing field.The key processes of geometric correction technology, image registration, image fusion research in the process of image mosaic during the task of low-altitude UAV remote sensing images from the carried digital camera have been researched in this paper, the concrete contents in this paper are as follows:First, according to the remote sensing image geometric distortion created by the interference factors during UAV tasks, terrain condition and atmospheric conditions, polynomial algorithm is introduced for geometric correction. Using ground control points for positioning firstly, space converting of coordinate system for remote sensing image. According to the type of geometric distortion. Setup coordinate system between a two unknowns variable polynomial of degree N so that the corrected coordinates can be gotten. Analysis of the advantages and disadvantages between nearest neighbor interpolation algorithm, bilinear interpolation algorithm and three cubic convolution interpolation algorithm, choose nearest neighbor interpolation algorithm in pixel gray resampling the corrected image. At the end, through image processing experiments of the images to mosaic which not be corrected and corrected, the experimental results show that the effect of geometric corrected images will be good at image mosaic.Second,according to the defects of SIFT high computational complexity, long time consumption and so on, transferred Gauss two order differential template and convolution operation of image function to the integral image subtraction of Box filter.At the same time, the dimension of feature vectors of feature detection operator was decreased through introducing SURF operator. The experimental results showed that improved SIFT algorithm increased the operation speed with the same robustness andmatching rate as original SIFT algorithm’s.Third, during the image fusion processing, the weighted average fusion algorithm of image fusion’s key technology has been analyzed, because the weighted average fusion algorithm can only fuse to overlap with rectangular area, however, the UAV remote sensing image with irregular features overlap region. So the traditional weighted average fusion algorithm has been improved in this paper with weights automatically match the irregular polygons, for the purpose of elimination of stitching marks. Through the image processing and analysis experimental comparison showed that the improved fusion algorithm can make the transition area smooth integration,can eliminate the image stitching line.