Accurate localization of circular mark based on interpolation algorithm and modified Hough transform

The recognition and precise centering of circular fiducial mark play a critical role in the localization of printed circuit board (PCB) and electronic components mounting. This paper proposes an effective and practical approach to detect the circular mark and calculate its accurate position in the image. The basic idea is to first estimate the center and the radius from the segmentation region of circular mark, then generate a high resolution image based on the edge-directed spatial interpolation from the original intercepted image only containing the mark and small background regions and finally calculate the accurate center coordinate of circular mark using the modified Hough transform. The novel hybrid scheme has improved the positional accuracy of the circular mark to approximate 2.5 mum which is equal to a quarter of the side of the pixel in the original captured image. Experimental results demonstrate that the approach has achieved noticeably better performance than the fast detection method in the precise centering.