絲餅瑕疵的自動光學檢測

Abstract

紡織產業一直是台灣重要的傳統產業之一，近年來更由於機能性紡織品的需求漸高，連帶提昇台灣紡織產業的價值。絲餅是一種由化學纖維紡紗工廠生產，提供給下游織造工廠織造布匹的原料，目前先進的紡紗工廠已可大量使用自動化生產設備來生產絲餅，但唯有瑕疵檢測這個步驟，仍需仰賴傳統人力目測檢視。為了降低人力成本，改善人工目視檢測難以標準化的缺陷，進而提昇檢測品質，我們提出將自動光學檢測流程應用於絲餅瑕疵檢測的方法。本論文針對絲餅的三種外型瑕疵進行瑕疵檢測，此三種瑕疵分別為：外型變異瑕疵、側凸瑕疵以及凸肩瑕疵。檢測流程中我們首先會校正攝影機的位置與光軸指向，確保拍攝到的影像品質，然後在正式檢測瑕疵時，利用受測絲餅影像上的邊緣資訊與正常絲餅應有的邊緣資訊數值進行比較，判斷該影像是否有外型瑕疵。最後本系統會儲存檢測過的絲餅各項參數，以作為將來建立資料庫時的資料來源。

The textile industry has always been one of Taiwan's important traditional industries. In recent years, the demand for functional textiles has been increasing, which has brought about the value of Taiwan's textile industry. The bobbins are the raw material of the fabric, and those bobbins produced by the spinning factory are made by automated processes, except the defect detection operations. In order to reduce labor costs and improve the quality of defect detection, we propose a method to apply the automated optical inspection process to the defect detection of bobbins. In this paper, we are dealing with three kinds of flaws of the bobbin. They are: Abnormal shape, Saddle, and Bulge. First, we will do the camera position and optical axis correction. And then in the formal defect detection processes, we will compare the edge information of the bobbin with normal information, decide if the bobbin has defects. Finally, the system will collect the information of bobbins for future's database usage.