CMOS Sensor自動光學檢測機台之設計與開發

Abstract

影像感測器的功能是將光的影像訊號，轉換成電的影像訊號，一般可分為電耦合元件(CCD)技術和互補金氧半導體(CMOS)技術兩種。CCD技術大部分由日本廠商主導，近年來由於台灣半導體技術成熟，使得CMOS影像感測器在台灣發展的非常迅速，應用範圍也越來越廣。目前CMOS影像感測器已成為很多商業化產品的主要元件，故CMOS影像感測器生產廠商對於如何確保經由封裝生產製程後，能有道關卡剔除瑕疵品，視為很重要的課題。CMOS影像感測器主要會發生之瑕疵包括感光區異常(含變色、異物附著、刮傷)、封蓋溢膠、封蓋偏移、及玻璃異常(異物附著、刮傷)、基板(Substrate)異常…等，這些瑕疵都會造成最終產品成像出現缺陷。

目前業界多是以人工目視對CMOS影像感測器進行全檢，受限於人眼的最小可辨識能力約20μm，故須借助顯微鏡進行檢驗，容易造成視覺疲勞而影響檢驗成效，若有不良品流出即會造成商譽的損失。本研究利用數位影像系統設計適當光源與取像機構，配合自行開發之AOI(Automatic optical inspection)軟體演算法，架構出完整的自動視覺檢測系統。

本研究的目的，在發展出一套適用於CMOS 影像感測器自動視覺檢測的機台，使能檢測出CMOS 影像感測器所可能會發生的瑕疵，以取代目前之人工目視檢測作業，改善目前人工檢測所不足的檢測精度，並提昇檢測速度與減少誤判率，最終目標達到減少人工成本與維持產品品質的理想目標。

The CMOS image sensors have become the key component in various commercialized products. For CMOS image sensor makers, how to eliminate defective products at checkpoint after packaging process is a very important issue. The major manufacturing defects on CMOS sensors include：abnormality in photosensitive area (color change, foreign matter attaches, scratches), lid-sealing glue overflow, lid-sealing displacement, and glass abnormal (foreign matter attaches, scratches), substrate abnormal, etc. These defects will become flaws on finishing goods.

Currently industry hires human inspector to conduct the inspection for CMOS sensor. However, due to the natural limitation, human eyes’ inspection capability is approximately 20μm only. People have to rely on microscope to carry on the inspection. This will also cause vision exhaustion and affect the inspection result. This research is to design a suitable light source and image extraction equipment, and a set of (automatic optical inspection) AOI software algorithms to construct a fully AOI system.

It is to help current human inspection, improve insufficient examination precision of present manual inspection, increase inspection speed, and finally to reduce manufacturing cost and maintain high product quality.