Research And Implementation Of UAV Image Mosaic Technology

In recent years,with the rise of UAV aerial photography,and UAV is widely used in many fields due to its small mass,flexibility and convenience,such as military exploration,environmental testing,disaster relief and so on.However,there are also some limitations,for instance,the image size is small and single image is difficult to reflect on the scene.In this case,it need to splice several UAV images into a large scene panorama.Therefore,the research work on UAV image mosaic is of great significance.Due to the shortcomings of traditional image stitching algorithms such as low efficiency and large amount of calculation,it is unable to meet the requirements for drone image stitching effectiveness.This paper improves the image registration algorithm based on SIFT(Scale Invariant Feature Transform)and weighted average fusion algorithm,optimizes the splicing process and improves the efficiency of image stitching,which better meets the requirements of image vision.Firstly,the paper address the problem of the distortion of UAV image due to the weather,illumination and noise through pretreatment of the original image.It mainly includes three parts:image geometric correction,image enhancement and image smoothing.By using the technology of pretreatment,the distortion of the image is corrected,and the noise of the image is removed,and the visual effect of the image is improved,and a good foundation is laid for the post-image splicing.Secondly,the paper determine the coincidence area of the images to be matched by using the flight data of the UAV,the appropriate algorithms are selected for feature extraction and feature matching in these areas to complete the whole process of image registration.The main research features extraction algorithms are Harris algorithm,SURF algorithm and SIFT algorithm,and matching algorithm has two parts:the rough matching and fine matching.Based on the requirements of timeliness,this paper introduces a partition algorithm.It reduce the number of feature points extracted and the feature extraction time,improve the efficiency of image registration.Thirdly,the paper analyze four common image fusion algorithms,including the method of direct average fusion,weighted average fusion,wavelet transform and multi-resolution fusion.In addition,modify the method of the weighted average fusion to eliminate the problem that the splice gap and the brightness difference are easily generated in the splicing process,it also makes the fused image smoother than before.According to the above research results,the functions of pretreatment,feature extraction,feature matching and image fusion of UAV are realized by MATLAB programming.The whole splicing process is stable and reliable,and we obtain the expected experimental results in the end.