The Research Of Skeleton Extraction And Description Methods In Image Recognition

Image recognition technology based on skeleton will be one of the important directions in the future machine vision study.In view of the difficult issues in the field, some new thoughts and methods have been proposed and studied in the paper to solve the keys about skeleton extraction,description,and so on.In 2D binary images,with the combination of cognition theory in skeleton extraction,the paper obtained the object's visual main skeleton through curve evolution and optimized the medial axis effect of the skeleton.To solve problems of traditional skeleton description method,the paper has first introduced moment technology into skeleton description,and deduced the invariant moment formulas which satisfy the essential requirements of pattern recognition.Then, the paper improved the skeleton tree's establishment procedure and control rules,unified the topology similar measure and the shape similar measure,and thus established a set of methods for object similarity computation,which is characterized from coarse to fine, and top-down Pyramid's-like.Summing up the above,the paper launched the research which considers the skeleton as objeet's characteristic,and more enriched the technical system in image recognition domain.The experiment had proven the optimizing algorithm proposed in the paper and has achieved the anticipated design goal.