Research On Fast Splicing Technology Of UAV Remote Sensing Image Based On Feature Points

As an emerging aerial photography method,UAV remote sensing has the advantages of flexibility,high image resolution,low operation cost,and portability compared with satellite remote sensing.Therefore,UAV remote sensing technology has always been a hot research topic at home and abroad.Due to the limitation of the camera resolution and the flying height of the drone,the image range taken by the drone is small,and the amount of image information is not rich enough.In order to quickly obtain a panoramic image information,it is necessary to splice multiple small-scale images together to obtain a panoramic image.Aiming at these problems,this article has conducted in-depth research on image splicing at home and abroad,and proposed a fast UAV image splicing algorithm based on feature points.When the UAV remote sensing acquires image,as its light weight is easily affected by the high-altitude wind speed and the light intensity during shooting,the image needs to be pre-processed before image splicing to eliminate the obvious difference and image distortion caused by the image brightness to ensure the quality of subsequent image splicing.Aiming at the problems of traditional SIFT algorithm,such as time-consuming splicing and mismatching of feature points,an improved splicing algorithm is proposed in this paper after conducting in-depth research on the registration algorithm based on feature points,.First,in the construction and generation stage of the signature,the feature vector descriptor is reduced from 128 dimensions to 32 dimensions.At the same time,a circular window is used instead of a square window.Second,the K-D tree algorithm is used to search for the nearest neighbors to improve matching efficiency and the RANSAC algorithm is used purify the feature points.Finally,the weighted average fusion algorithm is used to complete the image fusion.This article evaluates the quality of image splicing through subjective evaluation and objective evaluation.And it selects three indicators of the objective evaluation,the peak signal-to-noise ratio(PSNR),splicing time and structure similarity(SSIM),to evaluate and analyze the improved algorithm results,which ensures the accuracy and validity of the experimental results.Finally,an image splicing software is designed and developed based on the MATLAB platform,which integrates the traditional splicing algorithm and the improved algorithm in this paper.The splicing software includes functions such as main operation selection,image display and objective evaluation parameters,and the splicing program has demonstrated good interactivity and operability through experimental tests.