Edge Detection Algorithm Based on Grey System Theory Combined with Directed Graph

Abstract

A new edge detection algorithm based on grey system theory combined with directed graph is proposed to improve the defects in traditional Canny operator profiled fiber contour extraction, such as the discontinuous edge and the false edge. Firstly, the strong edge as well as binary image is obtained respectively by grey prediction and Nib lack algorithm, and then the connected graph is emerged by edge linking algorithm. Combined it with the Regions of Interest (ROI) which are calculated by grey correlation degree, the edge of fiber is appeared. The flooding algorithm and contour tracking method is used to get the fiber's outer contour, based on it and ROI, the enhanced fiber image is computed. Finally, the edge detected algorithm found on directed graph is applied to extract the edge of fiber, and the complete edge is gained after the burr is eliminated. The experimental results show that the algorithm cannot only resist noise very well, but also determine complete edge, as well as improve the detection accuracy. Compared with Canny operator, it has better continuity and overcomes the phenomenon of double edge.