Line Intersection Method

 

For the fiducials like the ones shown in Figure 2(c)(d), we use line intersection method to obtain sub-pixel accuracy. For the shape as in Figure 2(c), the two short line segments in the same direction are fitted into one line equation. The other two line segments in the orthogonal direction are fitted into another line equation. These two line equations are intersected to calculate the center of the fiducial. The similar method can be used for the fiducial shape as in Figure 2(d).

Our proposed algorithm for achieving the automated segmentation and sub-pixel localization of fiducial marks is:

1. Obtain sub images that contain the fiducial marks near corners or edge-centers from the original image.
2. Select the colour band that best represents the colour of the fiducial marks if the images are colour images.
3. Perform attribute-based morphology and other related processing techniques to automatically segment the fiducials as described in Section 2.3.
4. Refine the positions of the fiducial marks obtained from the previous step by using the Optimum Location or Line Intersection method.