Papers by Keyword: Image Matching

Abstract: Image matching is a very important field of computer vision and image processing. In this paper we proposed a topology model and algorithm based on characteristic points. Firstly, detecting edge and points feature, then making topology model and proving the excellent properties of the model. Finally, using the proposed model in image mosaicking, at last the feasibility of the model and algorithm are confirmed.

Abstract: This paper is focused on camera calibration, image matching, both of which are the key issues in three-dimensional (3D) reconstruction. In terms of camera calibration firstly, we adopt the method based on the method proposed by Zhengyou Zhang. In addition to this, it is selective for us to deal with tangential distortion. In respect of image matching, we use the SIFT algorithm, which is invariant to image translation, scaling, rotation, and partially invariant to illumination changes and to affine or 3D projections. It performs well in the follow-up matching the corresponding points. Lastly, we perform 3D reconstruction of the surface of the target object. A Graphical User Interface is designed to help us to realize the key function of binocular stereo vision, with better visualization. Apparently, the entire GUI brings convenience to the follow-up work.

Abstract: In order to improve the robustness performance of SURF descriptor applied to stereo image matching, a new matching method is proposed. By using the ratio of minimum to second min Euclidean distance of corresponding features, we can get the coarse matching points aggregation. Then, the epipolar line is computed from calibration parameters. Correspondences are taken as correct correspondences, only if they fall into a small neighborhood of their epipolar line. Taken errors into account, the neighborhood is set (-3, 3). Using this restriction, we can get the correct matching points aggregation. The experimental results show that the correct matches and matching efficiency are better than RANSAC.

Abstract: In recently years, extracting images invariance features are gaining more attention in image matching field. Various types of methods have been used to match image successfully in a number of applications. But in mostly literatures, the rotation moment invariant properties of these invariants have not been studied widely. In this paper, we present a novel method based on Polar Harmonic Transforms (PHTs) which is consisted of a set of orthogonal projection bases to extract rotation moment invariant features. The experimental results show that the kernel computation of PHTs is simple and image features is extracted accurately in image matching. Hence polar harmonic transforms have provided a powerful tool for image matching.

Abstract: Relative orientation is one of the most important and classic tasks in photogrammetry. Gross error of the corresponding points may be caused by computer automatic image matching, the commonly used method is unsuitable to relative orientation. In this paper, a novel approach of stereo images relative orientation considering gross error was proposed. Firstly the initial elements of relative orientation are computed by direct relative orientation, then the refined elements of relative orientation are obtained by Huber and iterative algorithm in the continuous relative orientation, which can eliminate the negative influence of gross error. Experimental results show that the proposed approach can achieve more excellent performances in accuracy than the commonly used method in the corresponding points with gross error.

Abstract: In order to improve the localization capability of the robot, digital image matching which based on the frequency domain was adopted, the method could reduce the impact of environmental factors, and adopts the sub-pixel subdivision pixel matching technology through curve fitting, Experimental results show that this method can achieve accurate localization in manipulator assembly working .

Abstract: The paper analyze and improve the SIFT optimized algorithm, and proposes an image matching method for SIFT algorithm based on quasi Euclidean distance and KD-tree. Experiments show that this algorithm has matching more points, high matching accuracy, no repeated points and higher advantage of matching efficiency based on keeping the basic characteristics of SIFT algorithm unchanged, and provides precise matching point to generate precise image stitching and other related fields of the follow-up product. At the same time, this method was applied to the layout optimization and achieved good results.

Abstract: Because of the inherent conflict between visual area and resolution in rock microscope structure, during the study of the RCTS (Rock Core Thin Section) microstructure, we cannot focus on the multi-scale structure characteristics of the particles, pores and throats, and it is fail to satisfy the demands of a more comprehensive study. In order to solve this problem, a microscopic image stitching method in RCTS is proposed by applying SIFT (Scale Invariant Feature Transform) detection algorithm. This method can successfully solve the conflict between the visual area and resolution, overcoming the problem of inclining and deformation in images acquisition under the microscope and finally, achieving the seamless stitching of RCTS microscopic image for big visual area. The experimental results show that this method can improve the accuracy of rock analysis in microstructure and has important practical and theoretical significance for the development of tight sandstone reservoir.

Abstract: Image registration is one of the fundamental problems in digital image processing, which is a prerequisite and key step for further comprehensive analysis,considering the advantages of the algorithm in speed and its disadvantage of more false matching points,a image matching method based on RANSAC and surf isproposed.The experiments results show that compared with the other algorithms,the surf algorithm improves the matching speed，as well as the matching accuracy，and exhibits good performance in terms of resisting rotation，noise,and brightness changes．

Abstract: A matching method based on Canny operator and SIFT algorithm for binary image is proposed. The Canny operator was applied on source image to detect edge and generate binary image. SIFT algorithm was used to match and recognize the binary images. The results of experiment show that the method we proposed ,under the premise of exact recognition, not only hold the great mass of key information, but also reduce the volume of data in source image. This method has good application prospects in massive image search ,terrain matching and so on.