Research on Similarity Measurement of Linear Skeleton
|School||Huazhong University of Science and Technology|
|Course||Communication and Information System|
|Keywords||Skeleton Similarity measure Skeleton tree Multi-scale skeleton tree Topology marked vector|
Increasingly widespread and the degree of automation to further improve with the application of computer vision, pattern recognition problem has become a hot research topic. Its development of related technologies in the areas of virtual reality, industrial simulation, scientific computing, visualization, an urgent requirement. National Natural Science Foundation of China (No.60273099) - \In this paper, a skeleton tree the linear skeleton topology similarity measure algorithm, the linear skeleton shape similarity measure also made a preliminary study, this long-standing problem of matching skeleton, some research results. This paper first introduces the basic background of the definition and extraction of the skeleton, the similarity measure theoretical as well as the overview of the pattern recognition method. Then, the establishment of a novel skeleton tree model. Mapped to the skeleton in a tree, and the connection between the tree level and node mainly reflects the topological characteristics of the skeleton. For a general rule can not create a skeleton tree ring skeleton processing given the ring skeleton skeleton tree building process. , Establish a reasonable root node selected by a discussion of the various situations skeleton contraction by multi-scale continuous skeleton algorithm to establish multi-scale skeleton tree topology more complex objects, the topological characteristics of the objects reflected from coarse to fine; criteria laid the foundation for multiscale classification match. This paper presents a similarity measure algorithm based on linear skeleton topology of the skeleton tree. Eigenvectors of the adjacency matrix of the skeleton tree and the definition of topological marked vector, the distance the norm topology mark vector difference between the two on the match as the two skeleton tree nodes, constructed between the two skeleton tree node matching relations, and on the distance of the skeleton tree matching node distance is defined as the best match relationships. The topological distance between the skeleton from the skeleton of the tree represented by the matching distance, and use the size of the distance values ??as the linear skeleton topological similarity metrics. In the were lower computational complexity and time complexity of the case, the general two-dimensional graphics achieved a good result. The idea of ??multi-level matching of complex objects and propose the use of multi-scale skeleton tree so that the algorithm has broader applicability. Linear skeleton shape similarity measure also been studied using the skeleton of the largest inscribed circle radius, skeleton branches skeleton points shape information, the skeleton coordinate translation, rotation, stretching aligned operation transformation of the connection of the two endpoints of skeleton branches of the curve to the shaft forward, given the definition of the match the distance between the skeleton branches, and then establish the best match between the two skeleton skeleton branches, lt; WP = 4 gt; the skeleton shape matching distance is defined as the establishment of distance and match the best match between the skeleton branches. The shape-matching algorithm on a simple two-dimensional graphics and achieved satisfactory results. Finally, a summary of the full text and pointed out the direction of future research work.