Research On Key Algorithm’s Improvement Of UAV Remote Sensing Image Mosaic Technology

Image stitching is a technology that fuses two or more images of the same scene and merges them into a panoramic large-field image according to the overlapping parts.It is difficult to obtain a single large-field-view and high-definition image due to the limitations of the performance of existing imaging equipment in practical applications.In order to expand the observation range,image stitching technology needs to be applied to camera drones to improve Image acquisition capabilities.Image stitching is widely used in medical image processing,military,security and other fields with the development of science and technology.Traditional UAV aerial images are often too large,which leads to the problem of slow speed in feature matching.And the traditional algorithm has a high mismatch rate when matching feature points because of high-resolution of image.This thesis proposes an improved algorithm based on ORB features and spatial transformation models for stitching UAV remote sensing images to solve above problems.The main contents of this thesis are as follows:1.In order to solve the problem that the brightness of the image is significantly different or geometric distortion occurs due to the UAV shakes and the changes in shooting angle and lighting when the UAV acquires remote sensing images.This thesis researches image distortion correction,image denoising,image enhancement and other technologies to preprocess images to eliminate noise and distortion.And finally improve the robustness and accuracy of image feature extraction,matching,and fusion effectively.2.In image alignment stage,this thesis uses the FAST feature detection algorithm of the ORB algorithm and the Shi-Tomasi algorithm to obtain the image feature points quickly and accurately firstly.And then,characterizing the feature points by the ORB description algorithm.Finally,the feature points are matched fine after matched roughly.And precise image alignment will be done.3.Aiming at the problem that stitching and ghosting are likely to occur in the image overlap area after alignment.This thesis calculates the spatial transformation model of the image overlapping area,introduces a weight coefficient,and uses this model to replace the original spatial model of the SPHP algorithm.And use multi-resolution fusion algorithm performs image fusion to reduce ghosting of overlapping areas of images.Finally,this thesis has developed a set of integrated software system that obtains feature extraction,feature matching,and image stitching.Users can complete image stitching according to their actual needs.The interface of this software system is simple and easy to understand.The display of picture selection and stitching results is simple and clear,with good visual effects.