依據廠商需求設計一種運用高速線掃描取像裝置與高空間解析鏡頭之光機模組。配合影像處理與快速檢測演算法的開發，研發一種檢測驅動晶片瑕疵的光學檢測系統，以精確檢測晶背上大於規格的瑕疵。檢測系統由線陣列影像擷取、照明控制、瑕疵檢測及操作管理等模組構成，以檢測包含殘膠、缺角與晶裂等瑕疵。系統之取像照明採 LED 為光源，具有體積小、低功耗與高照度等優點。系統以低幾何形變之高性能線掃描光機取像及二值化邊緣像素統計分布方法，進行晶片瑕疵之偵測與尺寸解算。使得檢測系統具有：(1) 可在受檢晶片封裝的運送過程中取得檢測影像，避免停頓式取像的缺點；(2) 每一晶片對應一幅影像，不需影像幾何校正與拼接，可大幅縮減影像處理時間；(3) 以邊緣影像為基礎的新檢測法，可提高瑕疵檢測的精度與效率等優點。因此可大幅降低晶片瑕疵的檢測人力與時間，並且免用顯微鏡等檢測設備，具有可大幅降低成本及提高晶片品質控管等優勢。本系統已可整合於封裝設備之挑選機，並可依廠商之機台需求來修改及進行系統整合測試，以滿足驅動晶片瑕疵之自動化光學檢測需求。

According to customer requirements, an optical inspection system for driver IC defects on chip backside isdeveloped to identify defects with size of greater than criteria. An opto-mechanical device is completed to acquireraw images using a high line rate scanner and high resolution lens. A novel defect inspection algorithm is alsodeveloped for precisely defects detection. Proposed inspection system is composed of 4 modules including lineimage acquisition with 5  m spatial resolution, illumination environment control, defect inspection, and operationsand management to inspect defects including residual glue, chipping and crack. LED light source is used for imageacquisition because of small size, low power consumption and high illumination. System has some advantagesbased on a high line rate imager with low geometric distortion and a method of edge pixels statistic for defectsdetection and size derivation as follows: (1) acquiring chip raw images during packaging process to avoid the pauseimage acquisition; (2) each chip has a corresponding raw image then image calibration and mosaic can be removedto reduce processing time substantially; (3) inspection accuracy and efficiency can be enhanced based on chip edgeimage and novel detection method. Therefore, manpower, time and cost for manually defect inspection can besaved. And quality control can be improved comparing with the one by manual way.