Research On Image Mosaic Technology Based On SURF

Image mosaic technology has gradually become a research hotspot in machine vision,digital image processing and other fields,and has been widely used in industrial detection,medical image processing and military reconnaissance.At present,with the continuous development of transportation infrastructure,bridge health monitoring has been paid more and more attention.In order to reduce the difficulty of bridge crack identification and reduce the workload of bridge maintenance.This paper mainly focuses on the existing image mosaic theory and experimental analysis methods from three important aspects of image preprocessing,image matching,image fusion,image mosaic technology research,and point cloud registration method based on stereo vision to explore the depth information in three-dimensional mosaic.The main contents include:(1)Firstly,this paper introduces the research significance,research status and application fields of image mosaic technology,and analyses the application prospects of this technology in bridge cracks.Secondly,the key technologies of image mosaic are summarized.The basic flow of image mosaic,image acquisition,image preprocessing,geometric transformation model,image registration and image fusion are introduced.The key technologies of image registration and image fusion in image mosaic technology are mainly studied.The main methods and performances of image mosaic are analyzed and summarized.Existing problems.(2)In view of the fact that a single crack image can not present a complete crack in bridge crack detection and identification,this paper proposes a method of crack image mosaic based on SURF algorithm,which mainly uses SURF algorithm to extract and match feature points accurately,and designs three groups of experiments for SURF,SIFT and ORB under the conditions of light,noise and scale transformation respectively.Through comparative analysis,experiments show that SURF algorithm performs better in robustness and real-time performance under different conditions.This method can get more image features and useful information,and greatly reduce the follow-up workload.On this basis,the panoramic image mosaic method is studied.The panoramic image mosaic is realized by combining the image frame to the image frame.The result shows that the mosaic effect of panoramic image does not appear large distortion,and the SSIM value achieves high accuracy.(3)On the basis of two-dimensional,this paper discusses three-dimensional registration,introduces the basic flow of three-dimensional registration,analyses the basic principles of stereo matching and point cloud registration,and mainly uses ICP algorithm for stereo registration.In this paper,ICP registration method based on stereo vision is proposed,and the depth information extracted is analyzed experimentally.The results show that the data has high accuracy,which provides convenience for subsequent three-dimensional stitching and reconstruction.