自動化IC印字瑕疵檢測(Ⅰ)-字元辨識

Automated Inspection of IC Printed Mark Defect (Ⅰ)-Character Recognition

饒忻(Hsin Rau) ； 曾健維(Jiann-Wei Tzeng) ； 呂理邦(Li-Pang Lu)

《工業工程學刊》 20卷4期 (2003/07) Pp. 317-326

https://doi.org/10.29977/JCIIE.200307.0002

摘要

由於IC晶片構裝印字在字型上大多採用筆劃較單純且易於辨識之向量型文字，且IC印字辨識若要運用線上即時檢測，基於速度的考量，不能用運算太複雜的演算法，因此本研究以一個階層式的架構，透過特徵值擷取與編碼的方式包含節點數、水平垂直座標軸相似性、極值個數、節點方向性及節點四角定位等五項，來做IC晶片常用字型之阿拉伯數字及英文字母的混合辨認。在成效上針對字體大小介於41(上标 *)41 pixel至140(上标 *)140 pixel之間的18,000個字元樣本進行檢測，包括影像前處理等流程，平均辨認速度每字元0.03秒，字元辨識率達99.94%。

關鍵字

IC印字 ； 特徵值擷取 ； 特徵值編碼 ； 字元辨識

並列摘要

IC printed mark mostly uses less number of strokes and easily identified vector type of words in word type, and the IC printed mark recognition can not use much complicated algorithm as a result of taking into account speed. Therefore, this study uses a hierarchical structure by character extracting to recognize the Arabic numerals and English letters of the alphabet. As for the performance of this system, for a case of 18,000 character samples between 41(superscript *) 41 pixel to 140(superscript *) 140 pixel to recognize, including image pre-processing, the average of recognized speed is 0.03 second/character. The character recognized rate is99.94%.

並列關鍵字

IC printed mark ； feature extraction ； feature coding ； character recognition

參考文獻

Ahmed, P.,Suen, C. Y.(1987).Computer Recognition of Totally Unconstrained Handwritten Zip Codes.International Journal of Pattern Recognition and Artificial Intelligence.1

Ahuja, N.,Rosenfeld, A.(1978).A note on the use of second-order gray level statistics for threshold selection.IEEE Transactions on Systems, Man and Cybernetics.8(12)

Applied Intelligent Systems(1990).ValidMark Automated Print Inspection Application Overview.

Chang, M. C.,Chen, H. Y.,傅楸善 Fuh, Chiou-Shann(1997).第十屆電腦視覺、圖學、暨影像處理研討會論文集.台中:

Friemel, B. H.,Bohs, L. N.,Trahey, G. E.(1995).Proceedings of 1995 IEEE Ultrasonics Symposium, Vol. 2.