Research On Optimized Stitching Of UAV Image Based On SIFT Algorithm

Unmanned aerial vehicle (UAV) remote sensing image has the characteristics of good real-time, high resolution, low cost, etc, has become an integral part of the surveying and mapping in remote sensing, but the coverage of UAV image is small, need to use stitching algorithm to obtain the image of entire area range measurement. In image stitching process, feature extraction and matching is the core and key of the entire process, the selection of feature extraction and matching algorithm directly determines the image splicing quality and speed.Scale Invariant Feature Transform (SIFT) is currently one of the most active algorithm in the field of image matching, the algorithm has the characteristics of good stability, rich amount of information, good portability etc, and has been widely applied in the field of image registration. However traditional SIFT algorithm also has some disadvantages:Such as processing larger data volume images take longer, and for different splicing images using fixed feature matching threshold, would affect the effect of splicing. Through in-depth study of SIFT algorithm, aiming at the shortcomings of the SIFT algorithm, put forward a method of feature matching threshold adaptive computing in SIFT, in order to improve the quality of the feature points matching. At the same time, puts forward a method of image stitching based on the interested area of remote sensing, then narrow the scope of the retrieval of image feature points, reduce the amount of calculation of the algorithm, improve the speed of image stitching.In the realization of SIFT feature matching threshold adaptive computation, approach is first calculate the standard deviation of the ratio of minimum and second-smallest Euclidean distance of all feature points, and by fitting the experimental data, obtain the relation between the standard deviation and the optimal threshold. In the stitching of specific image, first calculate the standard deviation of the ratio of feature points’ Euclidean distance, then according to this formula to calculate the threshold of feature matching. By testing, the calculation results as threshold can get a good results.This article through to research on region of interest algorithm especially the GBVS algorithm, Put forward a method of remote sensing image stitching based on ROI, specific approach is firstly calculates the significant figure of the images, then extract the marked area, and extract SIFT feature points in the marked area, feature matching based on adaptive threshold value calculation and image stitching, the experiment results show that image ROI can be used to instead of the original image in matching processing, and this method can reduce computation and operation time of the algorithm, improve the speed of image stitching, get a better quality of stitching.