軟性電路板尺寸漲縮機械視覺分類系統研發

軟性電路板(Flexible Printed Circuits,FPC)組成材料PI(polyimide)具有受環境及外力產生伸縮的特性，時常造成製程中相關自動化設備無法全稼動生產，一般需要將FPC用二次元尺寸量測儀量测分類，但因二次元量测儀需將路徑走完一次才可完成量测一片FPC，速度緩慢造成產速降低及成本增加，故尺寸漲縮機械視覺（machine vision）分類系統研發是一重要課題。本論文以研究開發一套具有機械視覺分類系統的設備，藉以達到提高產品良率增加設備稼動率及降低成本等目標。此系統必須涵蓋機械視覺辨識分類及機械定位兩大系統。機械視覺辨識系統則是透過攝影機的CMOS (Complementary Metal-Oxide Semiconductor，互補性氧化金屬半導體)，配合光反射技術及USB介面針對影像中有興趣區域(Region of Interest，ROI)，利用影像比對(Pattern Match) 作特徵點學習及生產特徵點搜尋，利用影像邊緣偵測技術(Edge Detection )作圓量測設定並擷取被測物光學點影像，作視覺迴授至PC進行分類運算，輸出量測及分類到顯示系統進行顯示與紀錄。最後透過實驗證明，本系統可達到快速且精密分類的預期目標。

關鍵字

軟性電路板；機械視覺分類；影像比對

並列摘要

One important raw material for Flexible Printed Circuits (FPC) is polyimide (PI) film. The film itself can easily be influenced by the climate and external forces and which cause changes in dimensional stability. These special characteristics often affect the automated production equipments unable to produce consistently. Normally one piece of FPC needs to pass through a dual dimensional inspection system process completely in order to distinguish defects and therefore slows down the production process and increases costs. Therefore it is an important discussion topic for the development of a machine vision inspection system to maintain the PI film dimensional stability during the FPC process. This thesis discusses the development of a machine vision inspection system targeting many optimal objectives such as product quality, equipment utilization, and cost reduction. The system also needs to integrate both visual identification and mechanical positioning systems. The machine vision inspection system works through CMOS (Complementary Metal-Oxide Semiconductor) in its camera module together with light reflective technique and USB interface. It then conducts pattern matching analysis within the region of interest of the image field and runs a search thru the distinguished areas. In addition, the system utilizes image edge detection technology to set calculation specifications and acquire optical images of the object inspected. The system then finally feedbacks the algorithm data to PC for calculation and at last sends information results to monitor system for display, notifying areas to be segregated. Last of all, this optical identification system is proven to be precise and efficient to properly isolate potential defective areas of the PI film in the FPC.