Abstract: Aiming at the alignment problem of nontransparent template in microelectronic packaging, the optical design and alignment algorithm were researched, and the bidirectional machine vision alignment system was proposed. The sizes of mold plates were extended by setting up a transparent calibration plate parallel to mold plates to be aligned, and the upper and the lower mold plate were placed in different position for photograph. The problem that the camera could not be installed in narrow space and the metal mold plates were not perspective were overcome, which realized the indirect alignment, and an automatic alignment system was developed. Based on machine vision, the images were acquired through bidirectional vision system and the edge images of mold plate were extracted by the image matching algorithm and Canny operator respectively for the initial extraction and the second extraction at the subpixel level. Then the planimetric position error of mold plate pair was calculated indirectly through coordinate transformation, and the error was used to drive XYR workbench for alignment control. The experimental results show that the high precision automatic alignment of the non-transparent mold plates can be realized in practical operation environment.
Key words: machine vision; bidirectional vision system; automatic alignment