Abstract: In order to improve the speed and accuracy of vision-based feature point extraction for optical fibers in their precision alignment process,a rapid extraction algorithm of subpixel feature points for contours of optical fiber cores is proposed on the basis of an improved spatial moment algorithm.In the proposed algorithm,firstly,the initial position of optical fibers is quickly searched,and the pixel-level edges of fiber cores are extracted via tracking the cores and their end faces.Then,subpixel edges are located by means of an improved spatial moment method,and linear equations of the axle and the edges are derived after a linear fitting on the extracted subpixel edges.Experimental results show that the proposed extraction algorithm runs more rapidly and possesses higher precision and stronger noise robustness than the traditional method,and that it meets the requirements of real-time precision alignment of optical fibers.

Key words: computer vision, feature extraction, one-dimension spatial moment, optical fiber