紙托自動光學檢測系統之研究與評估

在紙托工廠業界裡，目前大多數都是以人工目測方式來進行紙塑容器各項瑕疵檢測，也因為經常性視覺疲勞，無法精確掌控檢測的一致性。隨著技術的發展，機器視覺的自動化光學檢測，對於自動化生產線來說是缺一不可，能取代人工目檢、縮短檢測處理時間，並且降低生產成本。本論文選定紙碗與紙杯作為研究對象，利用以非接觸式檢測的機器視覺方法，針對產品瑕疵檢測問題，總共有三項檢測方式，分別為內部檢測、外部檢測以及底部檢測，皆能檢測出髒汙、毛邊、破洞與薄厚不均瑕疵表面特徵。機器視覺檢測紙碗與紙杯平均所耗費的時間為0.3秒，相較於人工檢測約2.5秒的檢測時間，遠遠相差8.3倍的時間，擁有更大的產出表現。並且運用瑕疵畫素換算瑕疵面積，能夠符合人性客製化要求瑕疵踢料實際面積。

關鍵字

紙塑；機器視覺；自動光學檢測

並列摘要

In the pulp factory industry, the defects of the pulp container are detected manually at present, it is impossible to accurately control the consistency of the inspection for frequent visual fatigue.Automatic optical inspection of machine vision is critical to automatic production lines, which can replace manual inspection, shorten inspection processing time, and reduce production costs. In this paper, pulp bowls and pulp cups are our research objects, and the machine vision is used to solve the problem of product defect detection. Three detection methods: internal detection, external detection, and bottom detection, all of which can detect the surface features of dirt, burrs, holes, and uneven thickness.The average time for machine vision to detect pulp bowls and pulp cups is approximation 0.3 seconds, which is 8.3 times shorter than the inspection time of about 2.5 seconds for manual inspection.And the use of defective pixels from the machine vision to compare the area of measured defects, we can get more accurate defective area from the pixel domain.