200848721 九、發明說明 【發明所屬之技術領域】 本發明係有關於一種具有相片單元及投影片單元及改 良式穿透光照射裝置的消費性平台式掃描器，其係用以做 爲一種光電系統，可供在透明塑膠樣本之離線錯誤檢測中 同時進行光吸收性及光偏折性視覺缺陷的偵測及定量分析 。合適的透明塑膠樣本是例如聚甲基丙烯酸甲酯薄膜（ PMMA薄膜）、聚甲基丙烯酸甲酯模造件、聚碳酸酯薄膜 (PC薄膜）及由前述材料製成的複合薄膜。 【先前技術】 本發明係有關於以掃描器蓋部（1 3 )內設有投影片單 元（14)之消費性平台式掃描器（11)爲基礎而進行之透 明塑膠樣本（第2圖，編號1 5 )的離線錯誤檢測裝置的開 發。消費性平或式掃描器可以理解是指在一般電子零件中 央市場販售的平台式掃描器，例如Canonscan LIDE及相 容型號、Epson Perfection、Epson V10、Epson V100 及相 容型號或是Hewlett Packard的Scanjet型號者。這些掃描 器具有USB或Firewire連接裝置，用以連接個人電腦。 塑膠樣本（1 5 )是放在投影片單元（1 4 )及相片單元（1 2 )的感測區域內。藉由此種的感測/照射狀態’穿透光或 入射光可以偵測材料（1 5 )內的缺陷（1 6 )’其等可依大 小及型式加以分類。 利用設置於掃描器蓋部（5 )內的投影片單元（第3 -5- 200848721 圖編號6)的改良，平台式掃描器（1 )不僅可以用來在該 塑膠樣本（8 )內搜尋並分析光吸收性錯誤（9 )，亦可搜 尋並分析稱爲凝膠體的透明光偏折性錯誤（3 )。就此改 良而言，掃描器蓋部（5 )必須要在掃描器（1 )上設置成 可由附件（4 )加以向上升高。此外，投影片單元（5 )必 須要能以預定的方式移動，以使得感測器單元（2 )及穿 透光照射裝置（6 )離開他們原有的位置配置關係。線感 測器（2 )的整個觀測區域現在包含有投影片單元的非發 光區域（7 )。 習用技藝= 在塑膠的製造中，特別是在透明塑膠零件的情形中， 對於目視的純度有極大的要求。因此，在生產前，或是與 該進行中之程序同時，材料樣本要針對雜質及視覺缺陷做 離線的測試。其結果可提供有關於材料視覺純度的量化報 告。 就此目的而言，有許多有名的影像處理公司提供光電 系統；該等系統的說明可自下列聯結下載，例如：（例如 http://www.ocsgmbh/en/st4.php )。這些離線錯誤檢測裝 置係透過在一個不位在被感測之影像的景深範圍內的邊緣 上做不對稱照射而得以同時偵測光吸收性及光偏折性缺陷 。這些樣本測試器設有現代的照射及感測器單元，並可以 多種方式加以應用。 可商業取得的平台式掃描器僅係供用來將文字或影像 -6 - 200848721 資料（相片、幻燈片）加以數位化，但他們到目前爲止並 未被應用於其它目的來做爲錯誤檢測裝置，且亦沒有該改 良過的照射裝置。 工業環境中的樣本測試器的高水準性能要面對著購置 及操作複雜性上相對應的高水準財務投資。此外，對於簡 單的問題而言，這些系統通常是“過度設計”，因此並不 能合理化他們的購置及使用。 【發明內容】 問題及解答： 本發明是基於要開發一種以消費性平台式掃描器硬體 爲基礎的離線錯誤檢測系統，因之而提供適合於該問題的 便宜解決方案的目的。詳細地說，做爲一錯誤檢測裝置， 掃描器也必須要能偵測透明塑膠樣本上的透明之光偏折性 凝膠體。第3圖中所顯示出的本發明裝置可透過一種令人 訝異之簡單型式來達成這些目的。消費性掃描器在感測器 系統及光學系統的領域內是相當地成熟，高品質系統及他 們的應用領域並不必要侷限於文字及影像材料的數位化作 業上。 所需要的僅是一設有投影片單元（6 )的掃描器。在 此情形中，投影片單元所掃描的區域即限制了要加以檢驗 的區域。爲了要能在吸收性雜質以外，也能在樣本材料（ 8 )上偵測出透明的光偏折性凝膠體（3 )，必須要對感測 條件做簡單的改良。根據照射測試及範例性的工業系統所 200848721 得到的經驗，需要採用非對稱式照射 及一黑暗區域。 就此而言，光亮及黑暗區域間的 必須要位在感測器的景深（depth-of-爲該邊緣並不要被顯示出來。藉由蔣 置於掃描器（1 )上而可由附件（4 ) 成該一需求。因此，現在可以得到非 描器蓋部（5 )仍必須要在水平平行 定的方式移動。因此，在掃描作業的 光照射裝置（6 )及感測器單元（2 ) 定方式改變。這會使得感測出之影像 料中包含有一光吸收性錯誤（9 ) ， 及改良的投影片單元（6、7 )是位於 在感測器單元（線感測器）（2 )的 的光強度衰退。根據缺陷的大小，磨 素的信號會明顯地衰減。光偏折性鐘 ，現在會明顯地顯現出穿透光載器的 )及光發光部位（7 )間的邊緣狀況 器上相關的像素現在會顯現出較均勻 這些特徵區別讓光亮及黑暗缺陷與成 位之缺陷的凝膠體之間可以分類。在 的情形中，場景的解析度是由感測器 數量來加以決定的。此系統的景深亦 以最佳的方式加以檢驗，但並非唯一 ，亦即照射於一光亮 轉接部，即該邊緣， field )範圍之外，因 F掃描器蓋部（5 )設 加以升高，其可以達 對稱式的照射，但掃 於掃描的方向上依預 過程中，掃描器穿透 之間的位置也會依預 場景變暗。如果該材 ΪΪ該感測器單元（2 ) 該錯誤的區域內，則 相關像素內會有顯著 $測器單元（2 )的像 Μ吳，即凝膠體（3 ) J穿透光照射裝置（6 ，如同一透鏡。感測 背景爲高及暗的點。 爲具有光亮及黑暗部 該問題的該解決方案 .單元（2 )內的像素 決定樣本厚度，其可 的。 -8- 200848721 此掃描器是連接至一主機電腦上’以供進行資料處理 並根據型式及大小來對錯誤加以分類。 【實施方式】 本發明的實施 錯誤辯識系統的零件 本發明曾以Epson Perfection V 750做爲消費性平台 式掃描器（1 )來加以實施。 技術資料（取自資料表）： 具有DIN A4投影片單元的平台式掃描器 設定的物理解析度：6400 dpi * 9 60 0 dpi (以膠片固定器固定 樣本進行掃描） 光學密度：4.0 DMax 界面：USB2.0(B型插座連接器xl) 高速 IEEE 1 3 94 ( 6 針腳 XI ) 光電裝置：6線彩色矩陣式CCD，具有Epson微透鏡 113 280 像素的 6400 dpi ( 18880x2 條 x3 色） 光源：白色冷陰極螢光燈光線 此裝置是屬於消費性平台式掃描器之頂級價格範圍內 者’但相對於工業用解決方案仍可提供顯著的價格優勢（ 至1 00倍）。此裝置的高解析度對於價格而言是合宜的’ 這些製造商細節依可歸納出較低的實際價格。爲得到此裝 置的較佳解析度，樣本必須要放置於固定座上，但其是配 合於此問題的解決方案。 -9- 200848721 要加以檢驗雜質的樣本材料（8)是塑膠薄膜，具有 〇.3mm的厚度。在擠製之材料的情形中，本發明揭示該邊 緣方向與擠製方向必須要平行，以使得表面的不均勻不會 具有視覺上的作用，以確保最佳的錯誤辯識。 ί市ί田益附件（4 )是由例如p 1 e χ g 1 a s s ®製做的，但材料 並非重點。掃描器蓋部（5 )是設置成可由此附件（4 )加 以抬升5 cm ’並可依預定方式水平移動，直到具有所述特 性的凝膠體有利地顯示於感測到的影像內。 所用的掃描及影像編輯軟體是來自Lasersoft Imaging®的 SilverFast。 評估裝置一台具有USB2.0界面的普通個人電腦，其 內安裝有該分析軟體。 本發明的優點 相較於工業用樣本測試器，本發明方法提供重大的財 務優勢’並且是一種強力的錯誤檢測裝置，其可用於在透 明材料內找出不同的雜質。 照射裝置的修改並不需要改變現有的照射單元，也不 需要結合其他種類的照射方法。 此外，消費性掃描器是己知應用於家用及辦公室用的 ’因此在操作上相當簡單及直觀。 本發明之實施一般已知的非對稱式照射來偵測透明光 偏折性錯誤是可應用於任何具有投影片單元的平台式掃描 器上，即使該掃描器於穿透光載器的發光區域（6 )旁邊 -10- 200848721 不具有黑暗、非發光區域（7 )。在此情形中，本發明方 法可藉由在穿透光區域（6)內製做出一邊緣（第4圖， 編號7 )來加以實施。此非發光區域（7 )可利用例如黑色 膠帶來製做之。此系統的其他部位則相同於本發明。掃描 器（1 )的蓋部（5 )必須要設置成可由附件（4 )加以升 高’以供檢驗樣本（8 )上的吸收性（9 )及透明偏折性錯 誤（3 )。透過水平移動掃描器蓋部（5 )，可以在感測器 線（2 )與具有蓋部（7 )的投影片單元（6 )之間再次形 成用以偵測凝膠體（3 )的必要定位。 本發明亦提供使用三種照射配置來檢驗樣本的選擇。 他們是使用入射光、使用正常穿透光、以及使用根據本發 明方法的改良式穿透光來進行的檢驗作業。這些照射手段 可以個別使用或是合倂使用來進行錯誤檢測。 在PMMA薄膜上的凝膠體的情形中，所能偵測的錯誤 的最小尺寸是大約20μηι。 在ΡΜΜΑ薄膜上的視覺錯誤的情形中，所能偵測的錯 誤的最小尺寸是大約2 0 μ m。 範例= 範例1 : 聚甲基丙烯酸甲酯薄膜（PMMA薄膜）上的錯誤偵測 具有0.3mm厚度的聚甲基丙烯酸甲酯薄膜利用固定座 對正於掃描區域。改良式穿透光照射裝置係檢驗該材料上 的吸收性及偏折性錯誤。這些錯誤會因本發明方法而可觀 -11 - 200848721 察得到，並可根據型式及大小加以分類。聚甲基丙烯酸甲 酯薄膜上的“錯誤”可以理解是指裂痕顆粒、凝膠體、異 物材料內含物及表面錯誤。 範例2 : 行動電話顯示器之內含物的偵測 爲分析射出成型行動電話顯示器內的雜質大小’樣本 必須要放置於掃描器的感測區域內。樣本接著即以正常穿 透光檢驗之。此方法可提供有關於內含物的良好地形資訊 ，可供後續製做出有關於其尺寸的報告。 範例3 : 使用入射光進行內含物的偵測 爲分析擠製材料內的雜質，這些材料會以入射光加以 檢驗。所得到的影像包含有光吸收性雜質的錯誤對比最大 化。 供錯誤檢測裝置使用的OFIS_CR (離線錯誤檢測系統，由 Christian Rosner開發）軟體的說明 此OFIS_CR軟體實施影像處理鍊的步驟（參見第1 圖）。其亦與硬體及使用通信。此軟體是以IDL (互動式 資料語言）撰寫的。IDL是由 Research Systems，Inc. (RSI)這家公司生產的，並由 Creaso GmbH這家公司在德 國販售。透過使用IDL Virtual Machine ( IDLVM)，此程 -12- 200848721 式可以在各種平台上執行（免費下載·· http:/www.ittvis.com/download/chooseplatform.asp?urlPro ductCode = 13 )。爲能使用掃描器來進行錯誤檢測，掃描 器必須要連接至電腦上，而且必須要安裝控制感測作業的 軟體。掃描器的啓動需要有另外一種軟體的存在。在開發 過程中，其係使用來自 Lasersoft Imaging®的 SilverFast 掃描器軟體。掃描器之啓動的程式規劃及此軟體本身的影 像處理的選取並不是很方便的。 所得到的影像可使用此程式來加以檢驗，而使用者可 透過雜質參數來控制雜質的偵測及分類。在分類之後，此 程式可提供相關的結果及評估。第1圖顯示出此0 F I S _C R 依時序的執行過程。 所需的硬體及軟體 硬體： •具有USB聯結及一般周邊設備（監視器、滑鼠、鍵盤） 的個人電腦 •具有附件的Epson Perfection V 750 Pro平台式掃描器 軟體： • IDL Virtual Machine ( IDL VM) • LaserSoft Imaging® 的 SilverFast®掃描器軟體 在進行錯誤檢測之前，必須要先安裝軟體，同時連接 上硬體。 -13- 200848721 錯誤檢測作業是使用平台式掃描器及〇FIS. 來進行的。 此軟體與使用者間的通信是根據數値的輸入 壓按。 在軟體的啓動螢幕之後，使用者可以透過顯 話來選取要進行錯誤檢測的影像來源。 接下來，0FIS工作簿會以文字檔的型式顯 ，使用者可以進行所有有關於檢驗作業的資料（ 者名稱、樣本材料、製程相關數據）。其亦可以 的解析度（DPI標示）、連至掃描器軟體的路徑 瑕疵臨限値。這些數値在〇F 1C工作簿中是設定 。就此而言，這些標準値是系統開發過程中認爲 數値。 如果掃描器是設定爲取像介質，則影像的感 謂的掃描器軟體加以控制的。爲能同時偵測光吸 偏折性錯誤，必須要進行本專利中所提及的照射 影像前處理功能可以透過使用掃描器軟體來加以 樣本正在被掃描時，外部的程式會中止。 除了此功能以外，也可以自檔案內讀取影像 具有可讓舊的樣本以新的參數加以檢驗，而無需 次的掃描。 影像資料必須要TIF格式，以供進行錯誤檢 在檢查影像的灰階値上，使用能以互動的方 試的影像。 _CR軟體 及按鍵的 示出的對 現。在此 例如檢驗 改變影像 及此時的 爲標準値 是合理的 測是由所 收性及光 的改良。 進行。當 資料。這 進行新一 式查自測 -14 - 200848721 除了顯示在螢幕上的測試影像之後，也會顯示出目前 滑鼠位置的灰階値輪廓。就此而言，使用可以決定他是否 需要現有滑鼠位置的行或列顯示於灰階値內。 在有著前述的資訊後，使用必須要輸入灰階臨限値， 以供進行下一步驟的分類作業。材料的雜質會透過灰階値 之設定爲參數而偵測出來並加以分類。光亮臨限値、黑暗 臨限値及瑕疵臨限値會將任何的雜質精確地分類成一類。 在此情形中，光亮及黑暗臨限値是用來隔離背景雜訊，因 此僅有超過這些臨限値的數値會被視爲雜質。此一簡易分 類策略的基礎在於前面說明過的掃描器照射改良。 此一分類作業是依據下列規則進行的： •光亮錯誤：這些雜質是較感測之影像背景爲光亮者，因 此會超過光亮臨限値。 •瑕疵臨限値：這些雜質具有較背景雜訊遠小許多的灰階 値，因此會低於設定的黑暗臨限値。 •凝膠體：此雜質具有凝膠體的特性。某些偵測到的團塊 會較背景爲亮，而某些則較暗。但是，最小的灰階値不低 瑕疵臨限値。 •具有視覺效果的瑕疵顆粒：具有凝膠體特性的這些內含 物可藉由最小灰階値來與凝膠體區分出來。在瑕疵顆粒的 情形中，這是低於瑕疵臨限値。 在此例中，瑕疵臨限値是維持爲標準値，或是使用者 輸入至OFIS工作簿內的數値。在開發過程中，其發現到 -15- 200848721 ，內含物通常是低於灰階値3 0以下，而凝膠體則不會低 於這個數値。在設定分類參數時的方向上，中間影像列是 顯示爲灰階輪廓，因爲凝膠體特性僅能在掃描器感測時根 據邊緣方向而在列方向上觀察到。在同樣顯示出的測試影 像中，此列是標示爲白色。當有關於灰階的資訊在影像中 變動時，則顯示出平均灰階値及影像中灰階値的分佈。這 些臨限値在輸入至灰階値輪廓後會加總。 在錯誤檢測作業的此點之後，一結論對話會用來再次 輸入任一臨限値，以啓動或中止軟體。 如果分類參數組是用來開始錯誤檢測作業，則在分類 的過程中會顯示出雜質的數量及進度指示器。這可讓使用 者隨著時間來跟隨著分類的進行。 檢驗的結束是由聲音信號及資訊螢幕加以通知。 分類策略的要點如下ζ 前面已說明過的使用灰階臨限値的方法在利用入射光 或正常穿透光檢驗樣本時也可使用。由於這些照射技術是 用來偵測材料中的光吸收性錯誤，黑暗臨限値的設定在此 例中是很重要的。 軟體所得到的檔案 爲能根據尺寸已經量化的被偵測到的雜質，針對檢驗 的樣本品質做定量報告，這些資在檢驗後要由使用加以調 整。 除了檢驗過的測試影像以外，雜質的標記影像也會被 -16- 200848721 儲存。在此，每一雜質均有一編號。此編號可以供特別地 檢驗任何雜質的數値，並檢視個別的錯誤影像。 錯誤檢測所得的影像包含有偵測及分類的結果。此影 像中每一雜質均由一偵測框加以圍繞。這些圍繞著雜質的 偵測框會依據分類結果而有不同的顏色。此視覺資訊可提 供使用者快速檢視樣本品質。 “ 0 FI S所得工作簿”顯示出在錯誤檢測之前及期間 由使用者輸入的資料。此外，其顯示出不同雜質的數量。 對於每一瑕疵顆粒及凝膠體，會顯示出標示編號及尺寸評 估所得的結果。下面的特徵會加以考量：圓形等效直徑、 最大菲列直徑（Feret's diameter)、列方向上的雜質尺寸 。此工作簿會另外用來儲存測試時間。 “錯誤影像資料夾”會儲存每一種雜質類型的個別錯 誤影像。錯誤影像亦包含有雜質的尺寸資料。具有其所包 含的標示編號的檔案名稱可用來推論出測試影像中的錯誤 〇 爲處理凝膠體尺寸的評估上’這些會同樣地以nxm的 格式加以儲存。以此格式儲存的2 D數値矩陣（凝膠體尺 寸、各種尺寸的絕對頻率）可以載入至諸如Excel之類的 表格程式’以供後續處理。 這些相關的尺寸是以“評估圖”資料夾中的圖形形式 繪製，並可事先檢視。 錯誤訊息： -17- 200848721 在使用 OFIS. 者會由定義好的錯 【圖式簡單說明】 第1圖顯示出 第2圖顯示出 第3圖顯示出 第4圖顯示tt 置。 【主要元件符號說 1 :掃描器 2 =感測器單元 3 :凝膠體 4 :附件 5 :掃描器蓋部 6 :投影片單元 7 :非發光部位 8 :樣本 9 :錯誤 1 1 :消費性平台式 1 2 :相片單元 1 3 :掃描器蓋部 1 4 :投影片單元 _CR軟體進行錯誤檢測的過程中’使用 誤訊息加以告知錯誤輸入° 程式的邏輯示意圖。 掃描器的配置。 掃描器中的光源及光接收器的配置。 掃描器中的光源及光接收器的另一種配 明】 胃插器 -18- 200848721 1 5 :樣本 1 6 :瑕疵200848721 IX. INSTRUCTIONS OF THE INVENTION [Technical Field] The present invention relates to a consumer flatbed scanner having a photo unit and a slide unit and an improved penetrating light irradiation device, which is used as an electro-optical system It can be used for the detection and quantitative analysis of light absorption and optical deflection visual defects in the offline error detection of transparent plastic samples. Suitable transparent plastic samples are, for example, polymethyl methacrylate film (PMMA film), polymethyl methacrylate molded article, polycarbonate film (PC film), and a composite film made of the foregoing materials. [Prior Art] The present invention relates to a transparent plastic sample based on a consumer flatbed scanner (11) having a transparencies (14) in a scanner cover (13) (Fig. 2, No. 1 5) Development of an offline error detection device. Consumer flat-panel scanners can be understood as platform scanners sold in the general electronic parts market, such as Canonscan LIDE and compatible models, Epson Perfection, Epson V10, Epson V100 and compatible models or Hewlett Packard Scanjet model. These scanners have a USB or Firewire connection to connect to a personal computer. The plastic sample (15) is placed in the sensing area of the slide unit (14) and the photo unit (12). The defect (16) in the material (15) can be detected by such a sensing/irradiation state by penetrating light or incident light, which can be classified by size and type. The flatbed scanner (1) can be used not only to search within the plastic sample (8) but also with the improvement of the slide unit (No. 3-5-200848721, Figure 6) provided in the scanner cover (5). Analysis of light absorption errors (9) can also search for and analyze transparent light deflection errors called gels (3). In this regard, the scanner cover (5) must be placed on the scanner (1) so as to be raised upward by the attachment (4). In addition, the slide unit (5) must be movable in a predetermined manner to cause the sensor unit (2) and the transmissive illumination device (6) to leave their original positional configuration. The entire observation area of the line sensor (2) now contains the non-lighting area (7) of the slide unit. Conventional Skills = In the manufacture of plastics, especially in the case of transparent plastic parts, there is a great demand for visual purity. Therefore, material samples should be tested offline for impurities and visual defects before production or at the same time as the ongoing process. The results provide a quantitative report on the visual purity of the material. For this purpose, there are many well-known image processing companies that provide optoelectronic systems; descriptions of such systems can be downloaded from the following links, for example: (eg http://www.ocsgmbh/en/st4.php). These off-line error detection devices simultaneously detect light absorption and light deflection defects by performing asymmetric illumination on an edge that is not within the depth of field of the image being sensed. These sample testers feature modern illumination and sensor units and can be used in a variety of ways. Commercially available flatbed scanners are only used to digitize text or images-6-200848721 (photos, slides), but they have not been used for other purposes as error detection devices. There is also no such improved illumination device. The high level of performance of sample testers in industrial environments is subject to high levels of financial investment corresponding to acquisition and operational complexity. Moreover, for simple questions, these systems are often “over-engineered” and therefore cannot rationalize their purchase and use. SUMMARY OF THE INVENTION Problems and Answers: The present invention is based on the development of an off-line error detection system based on a consumer platform scanner hardware, thereby providing an inexpensive solution suitable for the problem. In detail, as an error detecting device, the scanner must also be able to detect a transparent light-deflecting gel on a transparent plastic sample. The device of the present invention shown in Figure 3 achieves these objectives through a surprisingly simple version. Consumer scanners are quite mature in the field of sensor systems and optical systems, and high-quality systems and their applications are not necessarily limited to the digitalization of text and imaging materials. All that is required is a scanner with a slide unit (6). In this case, the area scanned by the slide unit limits the area to be inspected. In order to be able to detect a transparent light-deflecting gel (3) on the sample material (8) in addition to the absorbing impurities, it is necessary to simply improve the sensing conditions. Based on the experience of the illumination test and the exemplary industrial system 200848721, asymmetric illumination and a dark area are required. In this regard, the depth of the sensor must be in the depth of field of the sensor (depth-of-the edge is not displayed. By placing the cam on the scanner (1), the attachment (4) Therefore, it is now possible to obtain that the non-scanning cover portion (5) still has to be moved in a horizontally parallel manner. Therefore, the light irradiation device (6) and the sensor unit (2) in the scanning operation are determined. The mode changes. This causes the sensed image material to contain a light absorption error (9), and the modified slide film unit (6, 7) is located in the sensor unit (line sensor) (2). The light intensity decays. According to the size of the defect, the signal of the wearer will be significantly attenuated. The light deflecting clock will now clearly show the edge condition between the light-transmitting part and the light-emitting part (7). The associated pixels now show a more uniform distinction between these features that allow for the classification of gels with bright and dark defects and defects in place. In the case of the scene, the resolution of the scene is determined by the number of sensors. The depth of field of this system is also tested in the best way, but not exclusively, that is, it is illuminated outside a bright transition, ie the edge, field, because the F scanner cover (5) is raised. It can reach the symmetrical illumination, but in the direction of the scan, the position between the scanner penetrations will be darkened according to the pre-scene. If the material is in the wrong area of the sensor unit (2), there will be a significant image of the detector unit (2) in the relevant pixel, that is, the gel (3) J penetrating light irradiation device (6, such as the same lens. Sensing background is high and dark points. This solution for the problem with bright and dark parts. The pixels in unit (2) determine the sample thickness, which is ok. -8- 200848721 The scanner is connected to a host computer for data processing and classifying errors according to the type and size. [Embodiment] Parts of the implementation error identification system of the present invention The present invention was based on Epson Perfection V 750 The consumer platform scanner (1) is implemented. Technical data (taken from the data sheet): Physical resolution of the flatbed scanner with DIN A4 slide unit: 6400 dpi * 9 60 0 dpi (fixed by film) Fixed sample for scanning) Optical density: 4.0 DMax interface: USB2.0 (B-type socket connector xl) High-speed IEEE 1 3 94 (6-pin XI) Optoelectronic device: 6-line color matrix CCD with Epson microlens 113 2806400 dpi (18880x2 x3 colors) Light source: White cold cathode fluorescent light line This device is among the top price range of consumer flatbed scanners' but still offers significant price advantages over industrial solutions (up to 100 times). The high resolution of this device is suitable for the price' These manufacturers' specifications can be summarized as lower actual prices. In order to get the better resolution of this device, the sample must be placed in Fixed seat, but it is a solution to this problem. -9- 200848721 The sample material (8) to be tested for impurities is a plastic film with a thickness of 〇3 mm. In the case of extruded materials, this The invention reveals that the edge direction and the extrusion direction must be parallel, so that the unevenness of the surface does not have a visual effect to ensure the best misidentification. ί ί 附件 附件 attachment (4) is by, for example, p 1 e χ g 1 ass® made, but the material is not important. The scanner cover (5) is set to be lifted by 5 cm from the attachment (4) and can be moved horizontally in a predetermined manner until it has The gel of the characteristics is advantageously displayed in the sensed image. The scanning and image editing software used is SilverFast from Lasersoft Imaging®. The evaluation device is a general PC with a USB2.0 interface. The analysis software. The advantages of the present invention provide a significant financial advantage over industrial sample testers' and are a powerful error detection device that can be used to find different impurities within a transparent material. The modification of the illumination device does not require changes to the existing illumination unit, nor does it need to be combined with other types of illumination methods. In addition, consumer scanners are known for use in homes and offices, and are therefore relatively simple and intuitive to operate. Asymmetric illumination, generally known in the practice of the present invention, for detecting transparent light deflection errors is applicable to any flatbed scanner having a slide unit, even if the scanner is in a light-emitting region that penetrates the light carrier (6) Side -10- 200848721 does not have dark, non-illuminated areas (7). In this case, the method of the present invention can be implemented by making an edge (Fig. 4, number 7) in the light-transmitting region (6). This non-light-emitting area (7) can be made using, for example, a black tape. Other parts of this system are identical to the present invention. The cover (5) of the scanner (1) must be set to be raised by the attachment (4) for inspection of the absorbency (9) and the transparent deflection error (3) on the sample (8). By horizontally moving the scanner cover (5), it is possible to form a re-formation for detecting the gel (3) between the sensor line (2) and the slide unit (6) having the cover (7). Positioning. The invention also provides for the selection of samples to be tested using three illumination configurations. They are inspection operations using incident light, using normal transmitted light, and improved transmitted light using the method according to the invention. These means of exposure can be used individually or in combination for error detection. In the case of a gel on a PMMA film, the smallest size of error that can be detected is about 20 μm. In the case of visual errors on the tantalum film, the minimum size of the error that can be detected is approximately 20 μm. Example = Example 1: Error detection on polymethyl methacrylate film (PMMA film) A polymethyl methacrylate film having a thickness of 0.3 mm was aligned with the scanning area by a mount. An improved penetrating light illumination device is used to test for absorbency and deflection errors on the material. These errors are observable by the method of the invention -11 - 200848721 and can be classified according to the type and size. The "wrong" on the polymethyl methacrylate film is understood to mean crack particles, gels, foreign material contents, and surface errors. Example 2: Detection of the contents of a mobile phone display To analyze the size of the impurities in the injection-molded mobile phone display, the sample must be placed in the sensing area of the scanner. The sample is then tested by normal light transmission. This method provides good topographical information about the inclusions for subsequent reports on their size. Example 3: Detection of inclusions using incident light To analyze impurities in extruded materials, these materials are examined with incident light. The resulting image contains an error contrast that maximizes the absorption of light absorbing impurities. Description of the OFIS_CR (offline error detection system, developed by Christian Rosner) software for error detection devices This OFIS_CR software implements the steps of the image processing chain (see Figure 1). It also communicates with hardware and usage. This software is written in IDL (Interactive Data Language). IDL is produced by Research Systems, Inc. (RSI) and sold by the company Creaso GmbH in Germany. By using IDL Virtual Machine (IDLVM), this -12-200848721 can be executed on a variety of platforms (free download http://www.ittvis.com/download/chooseplatform.asp?urlPro ductCode = 13). In order to be able to use the scanner for error detection, the scanner must be connected to the computer and the software that controls the sensing operation must be installed. The start of the scanner requires the presence of another type of software. During the development process, it used SilverFast scanner software from Lasersoft Imaging®. The programming of the scanner's startup and the selection of the image processing of the software itself are not very convenient. The resulting image can be tested using this program, and the user can control the detection and classification of impurities through impurity parameters. After classification, this program provides relevant results and assessments. Figure 1 shows the execution of this 0 F I S _C R in time series. Required hardware and software hardware: • PC with USB connection and general peripherals (monitor, mouse, keyboard) • Epson Perfection V 750 Pro platform scanner software with accessories: • IDL Virtual Machine ( IDL VM) • LaserSoft Imaging®'s SilverFast® scanner software must be installed with software before connecting to the hardware. -13- 200848721 Error detection jobs were performed using a flatbed scanner and 〇FIS. The communication between the software and the user is pressed according to the input of the number. After the software launch screen, the user can select the source of the image to be detected by error. Next, the 0FIS workbook will be displayed in the form of a text file. The user can perform all the information about the inspection work (name, sample material, process related data). It also has a resolution (DPI indication) and a path to the scanner software. These numbers are set in the 〇F 1C workbook. In this regard, these standards are considered to be in the process of system development. If the scanner is set to capture media, the image's sensed scanner software is controlled. In order to detect the light-off deflection error at the same time, the illumination image pre-processing function mentioned in this patent must be performed. By using the scanner software, when the sample is being scanned, the external program will be aborted. In addition to this feature, it is also possible to read images from the file with the ability to have old samples tested with new parameters without the need for multiple scans. Image data must be in TIF format for error detection. On the gray scale of the image, use an image that can be interactive. The alignment of the _CR software and buttons is shown. Here, for example, it is reasonable to test the change of the image and the standard 此时 at this time. The measurement is improved by the acquired property and light. get on. When the information. This is a new way to check the self-test -14 - 200848721 In addition to the test image displayed on the screen, the grayscale outline of the current mouse position is also displayed. In this regard, a row or column that determines whether he or she needs an existing mouse position is displayed in the grayscale 値. After having the aforementioned information, the use must be entered with a grayscale threshold for the next step of the classification operation. Impurities in the material are detected and classified by the grayscale 设定 setting as a parameter. Brightness, Darkness, and Limitation will accurately classify any impurities into one category. In this case, the light and dark thresholds are used to isolate background noise, so only those numbers that exceed these thresholds are considered impurities. The basis of this simple classification strategy is the improved scanner illumination as described above. This classification is based on the following rules: • Brightness error: These impurities are brighter than the background of the more sensed image, and therefore exceed the light threshold. • On the limit: These impurities have a much smaller gray scale than the background noise and are therefore below the set dark threshold. • Gel: This impurity has the characteristics of a gel. Some detected blobs are brighter than the background, while others are darker. However, the smallest gray scale is not low. • Visually effective enamel particles: These inclusions with gel properties can be distinguished from the gel by a minimum gray scale enthalpy. In the case of ruthenium particles, this is below the limit. In this case, the limit is maintained as a standard, or the number entered by the user into the OFIS workbook. During the development process, it was found that -15-200848721, the content is usually below the gray scale 値30, and the gel is not lower than this number. In the direction in which the classification parameters are set, the intermediate image column is displayed as a grayscale contour because the gel properties can only be observed in the column direction according to the edge direction when the scanner senses. In the test image also shown, this column is marked in white. When there is information about the grayscale changes in the image, the average grayscale 値 and the distribution of the grayscale 値 in the image are displayed. These thresholds are added after they are input to the grayscale outline. After this point in the error detection job, a conclusion dialog will be used to enter any threshold again to start or stop the software. If the classification parameter set is used to start an error detection job, the amount of impurities and the progress indicator are displayed during the classification process. This allows the user to follow the classification over time. The end of the test is notified by the sound signal and the information screen. The main points of the classification strategy are as follows: The method of using the grayscale threshold 前面 described above can also be used when testing samples with incident light or normal transmitted light. Since these illumination techniques are used to detect light absorption errors in materials, the setting of dark thresholds is important in this case. The file obtained by the software is a quantitative report on the quality of the sample that has been quantified based on the size that has been quantified. These resources are adjusted for use after inspection. In addition to the tested test images, the image of the impurity is also stored in -16-200848721. Here, each impurity has a number. This number can be used to specifically check the number of any impurities and to view individual erroneous images. The image detected by the error detection contains the results of detection and classification. Each impurity in this image is surrounded by a detection frame. These detection frames around the impurities will have different colors depending on the classification result. This visual information provides the user with a quick view of the sample quality. The “0 FI S Workbook” shows the data entered by the user before and during the error detection. In addition, it shows the amount of different impurities. For each particle and gel, the results of the label number and size evaluation are shown. The following characteristics are considered: circular equivalent diameter, maximum Feret's diameter, and impurity size in the column direction. This workbook is additionally used to store test time. The “Error Image Folder” stores individual error images for each type of impurity. The error image also contains the size information of the impurity. The file name with the label number it contains can be used to deduce the error in the test image. 〇 To evaluate the size of the gel, these are equally stored in nxm format. The 2 D number matrix (gel size, absolute frequency of various sizes) stored in this format can be loaded into a spreadsheet program such as Excel for subsequent processing. These related dimensions are drawn in the form of a graphic in the “Evaluation Chart” folder and can be viewed in advance. Error message: -17- 200848721 In the use of OFIS. The error will be defined [Simplified illustration] Figure 1 shows Figure 2 shows Figure 3 shows Figure 4 shows tt. [Main component symbol 1: Scanner 2 = Sensor unit 3: Gel 4: Attachment 5: Scanner cover 6: Slide unit 7: Non-illuminated part 8: Sample 9: Error 1 1 : Consumerism Platform type 1 2 : Photo unit 1 3 : Scanner cover part 1 4 : The schematic diagram of the program that uses the error message to inform the error input during the error detection of the slide unit _CR software. Scanner configuration. The configuration of the light source and light receiver in the scanner. Another specification for the light source and the light receiver in the scanner] Gastrointestinal -18- 200848721 1 5 : Sample 1 6 :瑕疵