Automatic Detection and Location for The Fiducial Marks and Reference Fiducial Marks
Muh-Don Hsiao, Chuen‐Horng Lin, Jr-Wei Chen · International Journal of Robotics and Mechatronics · 2014
Automatic detection and location are proposed for the fiducial marks (FM) and reference fiducial marks (RFM) for the LED wafer images in this paper. The LED wafer image has two FMs, where the obvious FM is in the upper layer and the obscure FM is in the lower layer. The upper RFM is automatically detected and the lower RFM is determined after using an image enhancement technique. The automated search FMs of LED wafer images are divided into four steps: rough search, FM matching, fine search and trimming for sub-pixel images. FM matching includes the sum of absolute differences (SAD) and boundary feature matching (BFM). There are two types of experiments on an LED wafer image. In the first type, the upper and lower FMs of an image are circular. In the second type, the upper FM of an image is crossed, while the lower FM of an image is circular. The result shows that if the difference between RFM and FM grey-scale value distributions is small, the SAD matching has good effect. However, if the difference is large, the BFM matching is better. The positioning trimming for sub-pixels has a better effect on low resolution images. To validate the effect of the proposed method, the results are compared with the results of manual image matching in this study. By comparing the experimental results with those of the manual measurement method, we find that there are smaller error levels and a better effect.