Research on the extraction and classification of the concave point from fiber image

Identification of the different components from the mixed fiber depends on the parameter measurement of different fiber. As these components are always overlapped or intersected in a magnified fiber image obtained by microscope, the appropriate image processing technology is very important to realize the image segmentation. This paper firstly extracts the fiber image contour, then finds all corner points from the image contour. Accordingly all the concave points are found out from the corner ones by means of the Vector Triangle Area Method. On the other hand, the fiber skeleton is also extracted, the bifurcation points in the skeleton image are chosen, and two judge rules are designed to determine the public intersected line segment. Finally, the concave points are classified according to the corresponding relationship between the public intersected line segment and the concave points, which achieves the segmentation of overlapping fiber image successfully. The experiment results show that, the new algorithm can achieve the segmentation accurately, with a good segmentation effect. In general, the proposed algorithm can not only effectively solve the problem that the single fiber is hard to be segmented from the overlapping or adhesive fiber image, but also offer good test results.