| 摘要: | 本研究開發了一套自動光學瑕疵檢測(AOI)系統，針對鑄造金屬表面進行檢測，並以鑄造金屬製成的高爾夫球桿頭作為研究對象。首先，設計了一套光源配置，其中包括主光源與被動補光結構，本研究利用一個具有指向性的光源照射至檢測物上進行取像，並找出最佳的取像條件以實現最佳成像品質。影像處理採用Vision Assistant作為工具，結合形態學操作、拉普拉斯濾波、二值化分割、粒子濾波與瑕疵分析技術，以處理大量影像數據。此外，使用LabVIEW開發虛擬儀器 (VI)，打造快速且詳細的即時檢測介面，為鑄造廠提供精確的瑕疵資訊。 而硬體的部分，本研究構建了一套完整的取像系統及光型量測裝置。取像系統由九顆CMOS排列組成，各自滿足對應區域的拍攝條件，分析桿頭定位容許誤差範圍後，設計了配備快速夾治具的桿頭定位平台，簡化操作並確保定位的準確性。針對光源需求，透過實驗分析出鑄造桿頭表面瑕疵檢測的良好光照角度區間，同時，詳細量測與記錄光型、散射性、照射角度及照度，確保後續檢測中光照條件的得以重現。在檢測面積僅為 0.022mm² 的微小瑕疵時，系統達到漏篩率為零，也就是無出現漏篩的前提下得出九成左右的檢測成功率。為滿足鑄造廠的實際需求，系統整合了多種功能模組，如瑕疵數量分析、瑕疵分佈分析與大型瑕疵檢測，並建立分類系統，便於後續瑕疵處理或者人員複檢。檢測全程僅需約五秒，無需進行角度或光源調整，大幅降低操作人員手動失誤的風險。本系統有效降低光學瑕疵檢測成本，同時兼具高便捷性與功能性，為傳統鑄造廠提供高效的 AOI 檢測解決方案。;This study developed an automated optical defect detection system aimed at inspecting the surface of cast metal objects, specifically focusing on golf club heads made from cast metal as the research subject. Initially, a light source configuration was designed, comprising a primary light source and a passive auxiliary lighting structure. The study utilized a directional light source to illuminate the target object for image capture, identifying the optimal imaging conditions to achieve the best image quality. Vision Assistant was employed as the primary tool for image processing, incorporating morphological operations, Laplacian filtering, binary segmentation, particle filtering, and defect analysis techniques to handle large volumes of image data.Additionally, LabVIEW was used to develop a virtual instrument (VI) to create a fast and detailed real-time inspection interface, providing precise defect information for foundries. On the hardware side, the study constructed a complete imaging system and a light pattern measurement device. The imaging system consisted of nine CMOS sensors arranged to capture different regions and was equipped with a golf club head clamping and positioning platform, along with a Python-developed positioning verification program to simplify operation and ensure positioning accuracy. To meet the requirements for lighting conditions, the system meticulously measured and recorded the light pattern, scattering, illumination angle, and intensity to ensure the consistency of lighting conditions during subsequent inspections.When inspecting micro-defects with an area as small as 0.022mm², the system achieved a zero omission rate, maintaining a detection success rate of over 90% without any undetected defects. To meet the practical needs of foundries, the system integrated various functional modules, including defect count analysis, defect distribution analysis, and large defect detection, along with a classification system to facilitate subsequent defect handling or manual review. The entire inspection process takes approximately five seconds, requiring no adjustments to angles or light sources, significantly reducing the risk of human error during operation. This system effectively reduces the cost of optical defect detection while maintaining high convenience and functionality, providing an efficient AOI inspection solution for traditional foundries. |
