| A new method for earth observation, unmanned aerial vehicle, as known as UAV,remote sensing system has the advantages of flexible, convenient and efficientcompared with traditional satellite remote sensing system. UAV remote sensingsystem is used in fields such as land monitoring, resources exploration and disastersurvey. However, there are some obstructions such as the size of photos captured byUAV remote sensing system is small, the number of images captured from video islarge, and photo’s geometric deformation caused by airframe’s jitter and the overlapregions are often irregular graphics, the mentioned obstructions affect the result andefficiency of UAV remote sensing system’s image mosaic and its practical application.As a result, the new UAV remote sensing system’s image mosaic in real timeproposed in this research has important practical significance.In order to solved the problems of inefficiency, workload and others in thetraditional image mosaic method, the SIFT algorithm is improved and the mosaicprocess is optimized, which can save lots of time and energy. As a result, thedeveloped system is not only guaranteed the mosaic quality, but also satisfied therequirement of real-time system.First of all, because of the heavily deformation of remote sensing image, theoriginal image has been preprocessed by geometric correction without control pointand relative radiometric calibration, which can get rid of the restrictions of groundcontrol point and expand the range of system application.Secondly, the overlap region of two images is determined by using the UAV’sflight status parameters, and then some features of the mentioned overlap region canbe extracted, which can be completed the image registration. In this improved SIFTalgorithm, the extracted features are reduced and the feature extracting time isshortened largely at the same time, therefore the efficiency of image registration canbe improved significantly.Thirdly, the transformational matrix obtained by RANSAC algorithm can be usedto complete coordinate transformation. The image fusion is realized by using animproved weighted average fusion algorithm, which can be used to eliminate the seamof image and get a better result.Finally, a new real time UAV remote sensing image mosaic system is designed and programmed based on vs2008platform, which includes the video decoding,features extracting, features matching and image fusion. The mentioned system canplay video automatically and complete image mosaic at the same time, which has theadvantages of stable and reliable, better maneuverable and performance. |
PDF Full Text Request