Research On UAV Digital Image Matching Algorithm

With the rapid development of drone technology and sensor technology,drone photogrammetry technology has been widely used in surveying and mapping,disaster relief emergency,geological survey,power inspection,3D modeling,meteorological detection and other fields,and has become one of the hot topics at home and abroad.UAV(unmanned aerial vehicle)photogrammetry has advantages that cannot be surpassed by traditional photogrammetry.However,drones are light and have poor stability,resulting in irregular image overlapping,large rotation angles,and large distortions in the images.Faced with such images,many of the original effective matching methods have various problems.Some algorithms can no longer match the same name points.Some algorithms can match the same name point,but the number and accuracy of matching are greatly reduced.Therefore,on the basis of summarizing the existing image matching algorithms,it is very important to study the matching algorithm for UAV images.This paper discusses various UAV image matching algorithms.At the same time,it deeply studies the gray-scale LSIM algorithm and SIFT algorithm based on feature points and combines with the advantages and disadvantages of the operator,the SIFT algorithm is improved to make it more suitable for UAV images.The main research work and results of this paper on UAV image matching are as follows:(1)The paper focuses on the research of gray-based LSIM matching algorithm and feature-based SIFT matching algorithm.The principle and matching process of the two algorithms are introduced in detail.(2)In order to improve the matching rate of LSIM algorithm,LSIM is combined with Moravec operator,Harris operator and Forstner operator.Comparing and analyzing the two algorithms through experiments,it is concluded that the SIFT algorithm is more suitable for drone images with large distortion and large rotation angle.(3)According to the features of UVA images,this paper improves the traditional SIFT algorithm,expands the detection range of extreme points,gives the threshold of self-adaptation for different images,and modifies the description operator.The experiment was performed on the improved algorithm.The experimental results show that the feature point extraction,the number of matching pairs,and the matching time are all improved.