The Research And Application Of Distance Transformation Based Peeling Algorithm For Fiber Recognition

Automatic fiber recognition is a complicated research topic which is related to image processing,pattern recognition,and computer vision,etc. Due to inaccuracy and high time complexity of manual fiber recognition, the computer aided fiber recognition has gotten more and more attention. However,automatic fiber recognition is still a relatively complicated research and no research result has been applied successfully.This paper is a part of the research,which is sponsored by the Foundation of National Excellent Doctoral Dissertation of P.R.of China and the Scientific Research Foundation for the Returned Overseas Chinese Scholars,State Education Ministry,and is on the specific demands of Shanghai Entry-Exit Inspection and Quarantine Bureau on textile and fiber test.The automatic recognition of the natural fiber as well as shaped fiber is the key of the project.The project was passed the appraisal of the General Administration of Quality Supervision,Inspection and Quarantine of the P.R.of China in December 2007.To begin with,the automatic fiber recognition system pre-process the fiber cross-section of the microscopic image,followed by extractiong the single fiber and calculating fiber characteristics.Finally,different types of fibers can be recognized by using the Support Vector Machine(SVM) classifiers.The key technologies of the automatic fiber recognition system contain image pre-processing,fiber image segmentation,fiber characteristics extraction and fiber recognition.In this paper,we focus on determining and separating overlapped fibers.And skeleton extraction, which is one method of the fiber characteristics extraction,is also the main content.With the limits of techniques in making fiber cross-sectional samples, there are many overlapped fibers in fiber cross-sectional micrograph, which have embarrassed the fiber image segmentation.Thus fiber image segmentation,aiming to separate single fiber from the overlapped fibers as well as from image background,is the key for fiber recognition.In order to separate overlapped fibers accurately,the determining of the overlapped fiber is the first step.In this paper,the outline of multi-scale triangle area representation is studied,and then it is applied to detect overlapped fibers.The experimental results show that the algorithm can determine the overlapped fiber efficiently,provide accurate input for separate algorithm,and improve the efficiency of the fiber recognition system.The precision of segmentation directly change the extraction accuracy of fiber features.In this paper,by reviewing many image segmentation algorithms,we present a distance transform based peeling algorithm for fiber separation.With experimental results,the proposed algorithm can accurately separate the overlapped fibers,and the satisfactory result can be obtained for the separations of mass overlapped fibers.The skeleton is important for object representation and recognition. Skeleton-based representations are the abstraction of objects,which contain both shape features and topological structures of original objects. It can easily save as tree or graph structure which is convenient for object match.In this paper,based on the study of the existing skeleton extraction method,the distance transform based peeling algorithm for skeleton extraction is proposed.The algorithm can extract continuous skeleton to meet the human visual perception.The skeleton achieved can not only locate in accurate position,but also be stable for boundary noise.The amazing aspect of the skeleton is that it needs no pruning.This algorithm has proved to have high practicality and efficiency,as well as good discrimination in recognition among various types of shaped fibers.