印刷電路版自動檢測之參照建立

Abstract

本論文的主要目的是建立一個印刷電路版自動檢測系統的比對參照，將業界所採用的雷射製圖機DPF文字檔，轉換成標準無誤的PCB圖形檔，取代現今利用人工挑選PCB的方法。再將此圖形檔運用影像處理的方法， 產生特定格式的比對參照，方便日後進行各種的檢測工作。 在研究上分為三大步驟。第一，徹底剖析DPF文字碼，盡可能的將所有不同功能，用數學式子加以描述，製作成1:1的PCB數位影像。第二，將特定的圖形透過細線化與中斷點兩種不同的影像處理方法，比較兩者的優劣，並選取一個最適當的的方法作為我們最終處理的結果。最後則將結果製作成我們自訂的文字檔，此檔可以作為比對時的參照或是一個完整的資料庫。 在分析上，細線化我們採用Zhang and Suen's Method的方法，而中斷點則簡化k chain-code與 k differential chain-code之法，並結合與改良圖樣辨識中的A Distance Threshold Clustering Algorithm來進行中斷點的整合工作，並產生最後的比對參照。 由於PCB所要求的解析度非常的高，所以本論文試圖將多種影像處理的方法，在不影響結果的前提下加以改良簡化，以期能兼顧精確度與速度雙重要求，達到實用的最終目的。

This paper is to establish a reference of automatic PCB inspection system. The program is to automatically transfer the DPF text file; which is very popular used in the PCB industry and the input of laser ploter, to the standard and exact PCB digital image file. This is to replace the process of choice by hand. Then we use digital image processing to creat a specific reference to proceed all kinds of PCB inspection in the future. There are major steps in our reasch. The first, we analyze the DPF text file and do our best to defined it's various funtions by formula. The process will come up the 1:1 PCB digital image file. The Second, we took the specific graph through different image process ways, i.e. thinning and break point method. Then we picked up the better one to be our final output. Finally, we transfer the captioned output to a text file, which will be a reference data during comparison or a impeccable data base. In the analytical process, we use "Zhang and Suen's method" to thin our image, and use "K chain-code" and "K differential chain-code" to find break points. Finally we improve the method "A distance threshold clustering algorithm" in pattern recognition to arrange our break points, and establish final refernece. Becaues PCB requires very high resolution, this paper try to simplify many digital image processing methods wihtout changing result. At last we hope we can both have accuracy and speed to use in the PCB industry.