Research Of Matching And Stitching Method For Reconstruction Of3D Images With2D Image Sequences

3D reconstruction based on monocular2D image sequences is one of hotspots incomputer vision. The matching and Stitching of image sequences are the keytechnology in3D reconstruction. High constraint condition exists in matching andstitching of image sequences in spatial location. The methods and theory can have theuse for reference to two free images. Due to the particularity of image sequences, wecan get the depth of object in images by adding some conditions on existing matchingmethods and these depths can be used to3D reconstruction of the object. This researchwill study and explore the matching and stitching of image sequences and will get thespatial opposite depth of image object in some precision. The main works is:1.Introducing the existing methods and theory of2D images that contains featureextraction of image, feature matching of image, the methods of image matching basedon feature points and so on. Then we introduce the basic concept and method of stereomatching based on image sequences. We main introduce global and local algorithm andexpound some popular methods of the two kinds of methods, such as stereo matchingbased on adaptive weight, stereo matching based on image segmentation, stereomatching based on belief propagation and so on. Then we introduce the basic conceptand method of stereo matching based on image sequences. We main introduce globaland local algorithm and expound some popular methods of the two kinds of methods,such as stereo matching based on adaptive weight, stereo matching based on imagesegmentation, stereo matching based on belief propagation and so on.2.We propose a method based on best similarity of local geometric figure. It takesthe invariance of barycenter of geometric figure in image local matching as theoryevidence. It makes the most use of the features of image and geometric figure offeatures and the experiments have proved that our method can obtain more precisetransformation matrix.3.The disparity map of stereo matching impresses the opposite location thatprovides theory evidence for3D reconstruction of image object. So the precision ofmatching of images affects the opposite depth of object points. In order to obtain highmatching and disparity map and in the premise of matching speed, we propose a methodof stereo matching based on parallel estimation of distribution algorithm. It matches two images by estimation of distribution parallel and can get disparity map of highprecision.