Design And Implementation Of UAV Remote Sensing Image Stitching System

In recent years, unmanned aerial vehicle(UAV) are widely used in land dynamic monitoring, disaster monitoring, resource exploration survey and practical areas for its rapid response, high efficiency, and large scale high precision characteristics,and this technology has obvious advantages in a small area or areas with difficult flying to get access to high-resolution remote sensing image rapidly.But Unmanned aerial remote sensing also has its shortcomings,such as too many images but has small picture,compared with the conventional photogrammetry method.In order to have an intuitive understanding of the aerial parts of the images,it is necessary to mosaic, so the research is meaningful to develope a uav remote sensing image stitching system.This paper studies the main contents are as follows:First of all, the original image should be preprocessed, image distortion correction,image enhancement, image smoothing is needed for some inadequate quality of image.Then image registration link is necessary for the further research, through analysis and contrast the ORB algorithm, SURF algorithm experiment and SIFT algorithm in the image rotation, illumination changes and noise with advantages and disadvantages, it can be found that the SIFT algorithm in unmanned aerial vehicle has obvious advantages in image processing.Secondly in terms of image fusion, weighted average fusion method is selected by comparing the effect of stitching, and the best added sutures steps is added in this basis to eliminate the spell seam produced by image fusion and "ghost" phenomenon.Finally based on the above research results, the image preprocessing, image registration, image fusion, and other functions by using c + + programming language,thus a uav remote sensing image stitching system is designed and developed.