本論文主要目的是應用於雷射曝光機，在曝光前的對位系統精準度提升，其方式為利用機器視覺建構一套對位邏輯，針對因鑽孔製程造成靶孔的孔位精度或品質不良的影像，找尋最佳定位邏輯以提升曝光對位精準度。當製程中靶形變化過大時，通常採用人工方式做靶孔對位選擇再做曝光，使用手動選靶標除了造成產能降低以外，且其對位精準度會因人而異。本論文運用 hough 變換，結合邏輯判斷，讓製程中靶標的變異透過強健的邏輯，克服曝光前對位的不確定性，以及運用限定靶的尺寸範圍建構一套快速又準確的靶標確認方法。本論文先以實際 CCD 測試平台離線方式做測試，可直接移植到實際生產設備，以達到無縫接軌，並節省了大量的時間。本論文提出的相應解決辦法，對位精度由 93% 可提升達 95% 以上。;A concept to overcome PCB drilling process which causes exposure alignment accuracy issue is proposed in this thesis, the typical issue is that two circle image can be specifically exist as a target. The traditional algorithm cannot identify complicate targets and it will cause PCB solder and component sides to shift and affect yield rate. This study uses offline CCD test bench to detect the targets, and use Hough transform algorithm to judge result. After hundreds time of adjustment, the overall system is then transferred to the mass production system. It can save try and error time. Through this new algorithm, we can provide exposure machines with stable results and also reduce manual operations for target verification. The experiment results show the reorganization accuracy rate is above 95% by using the proposed.
