Research On Image Stitching Technology Of UAV Aerial Video

Image stitching technology is widely used in geological survey,remote sensing image processing,computer vision,medical image analysis,virtual reality and other fields.With the development of UAV applications,UAV aerial image stitching technology has become a research hotspot.This article has conducted an in-depth study on the splicing method of UAV aerial video images.The main contents are as follows:(1)In order to obtain high-quality video key frames,a key frame extraction algorithm for UAV aerial photography is designed.This method first divides the original UAV aerial video into sub-videos through the KLT(Kanade-Lucas-Tomasi)algorithm and the characteristics of the aerial belt of the UAV aerial photography.Then the sub-video collection is processed,using VSM(Video Stabilization Module),IQM(Image Quality Module)and SDM(Shooting Distance Module)modules to select high-quality key frames for subsequent image stitching.(2)Aiming at the characteristics of UAV aerial images,drawing on the idea of Auto Stitch image stitching method,the UAV image stitching algorithm is realized.In terms of feature point extraction of UAV aerial images,three feature extraction algorithms SIFT,SURF,and ORB are compared through experiments,and SURF is determined to be the best feature point extraction effect and efficiency.Aiming at the problem of cumulative error in the process of multi-image stitching,the LM algorithm is used to minimize the square sum projection error of the image,and the camera parameters of all images are globally optimized,so that the stitched image error is smaller and the drift problem caused by the stitching process is reduced.In image fusion,in view of the shortcoming that the traditional weighted fusion method will cause loss of high-frequency information near the stitching seam,a multi-band fusion strategy is adopted to ensure that the loss of high-frequency information is reduced in the fusion area,and the image transition is more natural.(3)Collected UAV aerial photography data set for testing and experiment of UAV aerial image stitching system,and realized a UAV aerial image stitching system.Experiments show that the proposed key frame extraction algorithm for UAV aerial video can ensure that the extracted key frames retain as much scene information as possible.Compared with the existing methods,the realized UAV aerial image stitching algorithm has a certain improvement in the details and smoothness of image stitching.