應用小波轉換技術於藥罐污點檢測之方法

Abstract

在微利時代的來臨與產品品質要求日益嚴苛之際，在提升產能的同時降低生產成本與節省人力並保持一定的品質水準本為當今製造業的一項重要的課題。在製造業中供應穩定品質的原物料，保持產品的一致性則在後道工程中可減少繁複非必要的檢測工作，進而簡化製造流程。 本論文以醫藥製罐業為例，其所生產的醫藥用罐。在外觀上是消費服用者對其醫藥產品的第一印象及信心指標；雖然其藥罐的外觀品質並不會影響到醫藥本身的品質，但表面的污損或任何製造上的瑕疵卻會明顯地影響到該藥服用者對該藥物實際的品質信心。因此醫藥用罐的表面污損（stain）、瑕疵檢測是醫藥製罐業將其列為一道重要檢測程序的原因。 小波分析是目前信號信處理領域的有效技術，如在特徵提取、信號檢測具有低熵性、多分辨|生、去相關性、選基靈活性等特性，使它成為在信號去噪領域內的有力工具。 論文中以小波轉換在影像處理上的應用技術為基礎，帶入醫藥製罐業生產做為研究案例，進行藥罐曲面上的污損檢測。在不損及原始圖像下以2D小波轉換（2D Wavelet Transform）進行去噪使藥罐表面上細微的凹凸平滑化與圖像質量縮減。再以1D小波轉換（1D Wavelet Transform）中的奇異點檢測方式取代傳統以圖樣比對(pattern matching)的比對方式突破影像閥值(threshold)的限制，在曲面上區分辨別出藥罐上的污損（stain）與細小凹陷刻度的不同。本研究將使用已在各業界實用領域已有相當基礎與實績的 NI（National Instruments）LabVIEW 應用軟體進行程式設計，預期開發一影像檢測方法，使本研究之成果能讓醫藥製罐業界實際進行應用。

In today’s world, the trend to seek superior quality and utmost profit has become an absolute priority.To increase production quantity while decreasing costs and labors becomes one single most important challenge in the field of product productions. The ability to provide stable quality in raw materials helps to maintain consistent quality in final products. This can greatly reduce complex or unnecessary testing procedures, hence, simplifying the production process as a whole. This thesis provides an example of such, targeting specifically at the bottling production of the pharmaceutical industry. A consumer perceives the first impression and confidence rating of a pharmaceutical product mainly from the bottling or packaging. Even though the bottling appearance does not actually affect the quality of a drug, any impurity or stain in the packaging can drastically impair a consumer’s confidence in the quality of the medicine. This is the reason why the bottling industry for pharmaceutical companies includes the evaluation for visible stain and impurity as one of their main testing procedures. Wavelet Transform is an excellent technique for signal analysis. It offers technologies to determine specified signal, to evaluate low entropy, to form multiresolution analysis, to extract from correlation analysis, and it providesextensive wavelet bases. All of which makes Wavelet Transform exceptional in signal De-noise. This thesis proposes a method applying Wavelet Transform to the testing of pharmaceutical bottles. Wavelet Transform gives advantages to the De-noise of bottle surface resolution. 2D Wavelet Transform can De-noise signal resolutions from coarse to smooth without damaging the original resolutions, while compressing the resolution. Furthermore,using the singular points function from 1D Wavelet Transform can replace the traditional pattern matching function and its threshold limitation. Such method can detect stain from the coarseness on the bottle surfacemuch more in-depth. Using the LabVIEW software from National Instruments,this thesis presents an integrated software program includingall of the Wavelet Transform functions mentioned above. The proposed program offers an innovativestrategy for the testing and analysis of bottle surface stain. It elevates the functionality and effectiveness of the bottle stain examination used by the current pharmaceutical industry.