Research On Line Matching Algorithm Under The Corresponding Points Constraint

Digital photogrammetry is widely applied to extract three-dimensional target geometry attribute information from two-dimensional digital images. Image matching is a key technology in digital photogrammetry, at the same time, it's the research hot spot in computer vision, image registration and object recognition. Line features are widespread in the artificial target and environment, it's an important characteristic of terrain reconstruction. In view of this, a corresponding points constraint line feature matching method is proposed. On the basis of obtaining the corresponding points and line features, according to the relationship of the corresponding points and line features to conduct the line feature matching. This thesis work is mainly reflected in the following aspects:(1) Feature points matching:Firstly, using adaptive filter to reduce noise, Wallis filter to enhance image details to complete the image preprocessing. Secondly, introducing the commonly feature points extraction algorithm, through experimental comparison, SUSAN operator is determined to extract feature points. Finally, utilizing moving Z-Plane constraint to match the feature points, experiments show that the method can obtain accurate and reliable matching results.(2) Line extraction:Analyzing and summarizing the existing line extraction method, introducing three existing typical method. Then putting forward line features extraction method based on chain code and line extraction algorithm of Freeman chain code priority. Through experimental analysis, the efficient Freeman chain code priority straight line extraction algorithm is adopted to extract line features.(3) Line matching:Proposing a line matching method under the corresponding points constraint, Firstly, according to the relationship between the corresponding points and line features to complete the initial line matching. Secondly, optimizing the initial matching results. Thirdly, confirming the corresponding endpoints by epipolar constraint, extending matching lines to get the same length. Finally, combining the overlapping corresponding lines, accomplishing the line matching.