Automatic segmentation of fiducial marks using attribute-based mathematical morphology

This paper described a fast and automatic method for segmenting fiducial ma rks in an image taken with a metric camera. These marks could be at the fou r corners, four middle sides, or inside an image (reseau marks). The segmen tation was realized by using attribute-based mathematical morphology techni ques. The attributes that we used in the morphology processing step were th e area, aspect ratio, and orientation of the best fitted ellipse of an obje ct. The algorithm took about 0.25 s in the automatic segmentation stage. Th e positions of these automatically segmented fiducial marks were then refin ed to subpixel accuracy. Dozens of real images with different shapes of fid ucial marks were tested and reliable results were obtained. Accuracy compar isons were made using our method and previously published methods. (C) 2001 SPIE and IS&T.