Research To Line Feature Based Relational Matching Of Organized Object's Imagery

Recently, with the rapid development of computer vision and photogrammetry, close-range photogrammetry is widely applied in the field of architecture, industry, engineering, archeology, biomedical and etc. It has advantages ofMatching is a core and difficult issue in casting 3D reconstruction based on image. Most of the industrial parts used for measure are made up of the rules geometric elements, such as lines, circles, etc., but with the lack of significant texture feature, point feature matching methods often fail. To solve this problem, in this thesis, line feature matching is used in this thesis. Firstly, we use the improved least square template matching to extract image line feature. Then establish the relationship between image line features and build the association graph description as a basis for segmentation by using the face image, and proposed to generate a surface topology-based algorithm. Finally, the relational matching based on linear feature by using the generated topological relations, which include two steps: base match and match transmission. The base match is to determine a set of the same name lines. The definition of the match transformisson is the match results'transformisson by use of the result of base match and topology. Finally, all of the same name point, line and pology features are obtained. This paper studies the relational match method based on line feature which greatly high speed, accurate precision, no-contact with targets, adaption to kinds of special circumstances, etc. Inspection, as an essential part of industrial production, yet with continuous improvement of manufacturing and processing, is required more and more strict demands. As a key part of quality control, inspection system of industrial parts plays a role in production quality and efficiency to some extent. Under the above issues, this thesis does certain research to 3D reconstruction of small industrial parts (sheet metals and castings) based on the theory of digital photogrammetry and computer vision. reduces the computation of image matching and improves the accuracy of image matching. In the Windows platform using Visual C++6.0 programming language developes algorithms related procedures; the correctness and effectiveness of the theory and methods presentes in this paper has been verified in the experiments. The relational match based on line feature enriches relevant theories and methods in photogrammetry feature extraction and matching, and has laid a good foundation for subsequent three-dimensional reconstruction and measurement.