Stitching Method For Reconstruction Of Three-dimensional Images With2D Image Sequences

In dynamic target tracking and characteristics matching based on image sequences,a lot of studies have been carried out at home and abroad, and considerable progresshave been made, yet most researchs are focused on the plane image fusion stitch, not yetrelated to the fusion stitch of the three-dimensional image reconstruction fromtwo-dimensional image sequences. Although the plane image mosaic fusion method canplay a certain reference for three-dimensional image stitching and fusion, but can not beused directly on stitching and fusion of three-dimensional images in different scales,and the technology and methods are very different. The previous method can be drawnon, but can not be applied directly. This paper has a depth research on stitching methodfor reconstruction of three-dimensional images with2D image sequences, and mainlycompleted the following work:1. On the basis of reading a large number of domestic and foreign relatedliteratures about stitching method for reconstruction of three-dimensional images with2D image sequences, summarized the research status.2. Interpreted the basic theory and commonly used algorithm on image stitchingand fusion technology. Basic processes of digital image stitching, including imageregistration algorithm, algorithms to eliminate false matching, and image fusionalgorithm, are presented. Then, the description of these algorithms are showed throughexperiments.3. A stitching method for distorted images based on SIFT feature matching wereproposed, throuth the understanding of the principle of image distortion correction,SIFT feature matching and RANSAC algorithm. The method first integrated thecorrection algorithm; then extract the SIFT feature on each image, simple featurematching, and improved RANSAC algorithm to eliminate the false matching on thebasis of simple feature matching; Finally, based on the proposed method, two sets ofexperiments were implemented, and by comparative experiments, analysis of theexperimental results, the proposed method was verified.4. A stitching method for repeating texture images based on the structuralcharacteristics of overlapping regions was proposed, on the basis of Harris cornerdetection theory. Firstly, Harris corner features were extracted on image sequence, and corner features were simple matched; then a polygon area was built based on the matchpoints in overlap region, match the structural features of the polygon area and determinethe width of the overlap region that prepared for stitching, to achieve an exact match.Finally, experiments were carried out based on the proposed method, experimentalresults were displayed and analysised at the same time.