表面黏著型發光二極體瑕疵檢測系統之研究

Abstract

近年來，表面黏著型(Surface-Mounted Device，SMD)發光二極體(Light Emitting Diode，LED)已被大量生產並應用於背光模組光源，但SMD LED封裝廠生產製造後仍以人工目視檢測可能產生之六項瑕疵，包含缺件(Missing Component)、側翻(Wrong Orientation)、極性反(Inverse Polarity)、缺口(Mouse Bite)、金線不良(Gold Wire Missing)，以及表面不良(Surface Defect)等。本研究研發出一套SMD LED瑕疵自動光學檢測系統，此系統包含硬體機構與檢測演算法。硬體光學取像機構安裝於卷帶包裝機上，可擷取並收集捲帶上三種類型LED影像。檢測演算法對收集之影像先進行影像前處理與旋轉校正。接著，使用檢測區域自動產生法(Inspection Regions Auto-Generation Method)切割各LED的檢測區域。針對切割出的檢測區域擷取其特徵值，並由所提出之方法進行檢測與瑕疵分析。透過本研究所開發之檢測系統，實驗結果顯示平均檢測率高達97%，單張LED影像檢測時間小於0.25秒。可正確且快速地檢測不同類型的LED瑕疵，並提高SMD LED產品之品質檢測成效。

Surface-mounted device (SMD) light emitting diodes (LEDs) are used in many different applications. However, some LED defects are unavoidable in large-volume fabrication and taping processes. These defects may include missing components, wrong orientation, inverse polarity, mouse bite, gold wire missing, and surface defect. To inspect the defects, LED packaging factories still rely on human inspector that leads to high cost, time consuming, and lack of consistent inspection results. This dissertation is to design and develop a defect inspection system for the three types of LEDs. The proposed system includes the software system that is to inspect the mentioned defects and hardware system that is used to capture the on-line images. Image preprocessing and normalization are firstly applied to the on-line images. The proposed inspection regions (IRs) auto-generation method is then used to segment the IRs of the three types of LED images. The inherent features on the IRs are extracted to inspect corresponding defects. Experimentations reveal that the successful inspection rate is up to 97% and the inspection time of per LED image is less than 0.25 second. The contribution is that the proposed method could provide reliable and effective inspection results of the three types of LEDs.