200823448 九、發明說明： 【發明所屬之技術領域】 本發明係-種自動化光學檢測系統及其方法，其係特糊於一種能針 對密封元件之上下表面之表面區域、内部、内外環面之表面區域做檢測之 自動化光學檢測系統及其方法。 【先前技術】200823448 IX. Description of the Invention: [Technical Field] The present invention relates to an automated optical inspection system and method thereof, which are specially adapted to a surface area, an inner surface, an inner and outer toroidal surface of a lower surface of a sealing element. An automated optical inspection system and method for detecting areas. [Prior Art]
溫度的改變而產生許多瑕疵 進行產品的抽檢。 密封件的製造品肢妓善’考驗著各家_對品奸控與核心技術 的掌握能力，不賴是，密封件賴造在射域賴財可能會因壓力與 而目前最後一道品質的管控卻還是使用人工 在現代化流水生產線後面，相的卻是數十甚至上百驗測工人，執 行品質的控制。此不合理的現象，在給讀增純大人工成本和管理成本 的同時，仍然無絲證1_的檢出率。當今企業_競爭，已經不允許那 怕是千分之-的雜存在。㈣元件需要進行微小尺相精確與快速的測 里’形狀祕配，顏色的職等。這些工個人錄本無法敎持續地進 行，除此之外，人眼總是會因疲勞而發生漏檢，或是因為主觀判斷的不同， 出現品管標準寬鬆不一的現象。 再者’由於PU材質具有半透明的特性，傳統人工檢測是透過強光照射 下’内部若有瑕疯（例如氣泡或異物），會使光線在密封元件内部形成多次 折射，瑕疵的形狀與位置將會顯現。但是這種方法並不理想，首先人眼長 時間在強光下進行檢測，很料造成傷害，再者強光亦耗電，因此這種既 傷人眼又耗電的檢測方式，有必要加以改進。更確切的說，對於密封元件 5 200823448 品質的管控，傳統上是透過訓練有素的作業員，透過肉眼進行瑕疫的找尋 與種類的靖。值得—提的是，人卫檢雜容易因為人員的疲勞或其他各 種目素，使得即便是輯有素的檢測人貞，也無法保證其檢測是萬無一失 ‘的，導致重要的品質管控關卡變的不可靠。另-方面，人工檢測的速度並 不-致而域本X高，在整體現代姻流水生產線下，這種品質的管制方The temperature changes to produce a lot of defects. The manufacture of seals is good for both people's ability to master the control of the product and the core technology. It is not bad that the seals are based on the pressure and the current quality control. Still using artificially behind the modern production line, the phase is dozens or even hundreds of inspection workers, performing quality control. This unreasonable phenomenon, while giving reading to increase the labor costs and management costs, still has no detection rate of 1_. Today's enterprises _ competition, has not allowed even the existence of a thousand. (4) The components need to be accurately and quickly measured by the tiny ruler. These personal records cannot be carried out continuously. In addition, the human eye will always be missed due to fatigue, or the quality control standards may be different due to different subjective judgments. Furthermore, because the PU material has a translucent nature, the traditional manual detection is to illuminate the interior with strong light (such as bubbles or foreign matter), which causes the light to form multiple refractions inside the sealing element. The location will appear. However, this method is not ideal. First of all, the human eye is tested under strong light for a long time, which is expected to cause damage. In addition, the glare also consumes electricity. Therefore, it is necessary to detect the method of injuring the human eye and consuming electricity. Improve. More specifically, for the quality control of the sealing element 5 200823448, it is traditionally through the well-trained operators to find and categorize the plague through the naked eye. It is worth mentioning that it is easy for people to check and check because of the fatigue of personnel or other various factors, so that even if it is a well-tested person, there is no guarantee that its detection will be foolproof, leading to important quality control. Unreliable. On the other hand, the speed of manual detection is not high, and the domain is high. Under the overall modern marriage production line, the regulator of this quality
Ik著產質需求的不斷提高，目前仰賴人工目檢的方式 已無法勝任品管的工作。 _ 鱗決傳統人卫檢測的缺失，檢測密封元件時係使用機器視覺為基礎 的檢測方法；請參照作者江新偵，，，應用機器視覺檢驗油封之尺寸及表面 —疵私立το智大學，工業工程研究所，碩士論文⑽…係針對密封 1件之上下表面之表面區域（如第一圖所示，標示B處）進行瑕絲像與 H而最為重要包含這個唇狀前端_環與外環表面，竟無著墨。在這 個唇喊面若產生瑕疲，譬如到傷、發泡，則使用最多密封元件的流體機 _ 齡、錢心錢祕纽，將無法正常運轉造成龐大的損失。未檢測之 轉的翻❻外環表面屬於曲面，由魏前技術對—個非平面的物體進 =取像時’係朗面掃描攝影機分次分區域來娜非平面的影像，既耗費 «又因為大範圍的曲面投影成二維平面時，影像必定會失真，因此使用 面掃描攝影機取像並非有效的做法，因此無法使用面掃描攝影機一次操取 整_二維的曲面。因此，可以想見此方式對於密封元件的品質上的把關， 乂傳4*人JL檢測更為精準’但尚非是最為精準的檢測結果。 為此，本發明提供-種自動化光學檢測系統及其方法，以提昇檢測結 6 200823448 果的精準率，改善上述缺失。 【發明内容】 ，本㈣之轉目的祕触—種自触辟岐封法 .可針賴元件上下表面之表面區域„環面之表面絲 測，以提升檢測結果的精準率。 仃私 本發明之另-目的在於提供一種自動化光學檢測系統及其方法，轉 • rpu#f之密封元件，_其無透明的特性，個紅外«光板做為 月^源，對PU材質之密封元件進行取像，改善原來_強光才可透過肉眼 稍微看出内部之弧面區域之瑕疵的缺點。 树明之再-目的在於提供—種自動化光學檢測系統及其方法，其係 檢測密封元件内外環表面，利用了線掃描攝影機的特性，-次擷取-條掃 描線的影像，並透職轉運動即可將樣本表面展開，制二維影像便於後 績的影像處理，藉此檢測㈣元件⑽環面之表面區域之瑕疫。 • 本發鴨提供—種自動化光學檢«統及其方法，個-第-面掃瞒 攝〜機對PU材料之讀元件之上下表面之表面㈣進行取像及檢测，檢測 /务見Λ雀封元件的上下表面之表面區域有瑕，疵，即停止檢測。密封元件 ^上下表面之表面區域檢測無瑕紐，再以-帛二面掃喊影麟PU材料 之㈣元件之_進行取像及檢測，取糾以紅外線鴨㈣元件，將稍 具透明特性❺PU材質密封元件之内部之弧面區域之瑕，疵， 譬如氣泡或是異 物’透過紅外線的穿透力完整地將密封元件之内部之弧面區域之瑕疵成像 出來’檢測後發現該密封元件的内部有瑕疵，即停止檢測。密封元件的内 200823448 。I瑕症後再以一線掃瞒攝影機對崎料之密封元件之内外環面之 表區:^之瑕錢仃取像與檢測，檢測密封元伽外環蚊表面區域，係 _ 了線掃蝴械的雖，—次娜—麟描_影像，並透職轉運 本表面m㈣二維影像便於賴的影像處理，藉綠測密封元 件内外％面之表祕域；檢峨發現該密封元件_外環面之表面區域有 瑕疲卩彳如未發賴㈣元件的内猶瑕賴，即為通過檢測 之該密封元件。 树明運關器視覺技_發—套自動化光學檢_統，取代傳統人 抽仏的方^。本發日柯防止人卫檢稱致贼品的驗，造成當具有瑕 疲的密封件朗於昂貴的空油壓流體機械時，產生不可彌補的嚴重損失。 底下藉由具體實施例配合所附的圖式詳加說明，當更容易瞭解本發明 之目的、技術内容、特點及其所達成之功效。 【實施方式】 明參知第-圖及第二圖，本發明係提供—種自動化光學檢測系統及其 方法主要在提供一套密封元件之瑕疵檢測系統，其包含表面檢測裝置、 内λ測裝置、環面檢測裝置，藉此對密封元件作全面性的瑕檢測。密 封疋件包括PU迫緊(PU-Packing)、油封(0il Seal)、墊圈（Gasket)或任一 %形密封το件之聚氨基甲酸酯（pu)材料的產品。密封元件之瑕疵檢測包含 上下表面之表面區域（第一圖所示，、内外環表面（第一圖所示，A)、 及内部（第一圖所示，C)等檢測區域。基於密封元件(待測物）非單一平面， 因此使用面掃描與線掃描攝影機根據取像部位來分站檢測，共可檢測出瑕 8 200823448 疵有七種包含變形、毛邊、未密合、氣 '、 一物 '刮傷、及發泡。步驟S10 之表面檢測、步謂支内部檢測、步侧之環面檢測之架構與細節， 及本發明讀奴_容，__錢_細你 【步驟S10表面檢測】 步驟之表面檢測的檢測流程如第三圖所示，主要檢測密封元件的上 下表面之表通域（第-圖麻，B)。錢進行步_及遞所示，使Ik is constantly improving the quality of production, and currently relying on manual visual inspection is no longer qualified for quality control. _ The scale is determined by the lack of traditional human health detection, the detection of sealing components is based on machine vision-based detection methods; please refer to the author Jiang Xinzhen,, the application of machine vision to test the size and surface of the oil seal - 疵 private το智 University, industry Engineering Research Institute, Master's thesis (10)...The surface area of the upper surface of the seal 1 (as shown in the first figure, marked B) is the most important and contains the lip front end _ ring and outer ring On the surface, there is no ink. If there is fatigue in this lip-screaming face, such as injury or foaming, the fluid machine with the most sealing elements will be unable to operate normally and cause huge losses. The surface of the untwisted turning outer ring belongs to the curved surface, and the image of the non-planar object is taken by the Wei-pre-technical method. When the image is taken, the image is divided into regions and the image is not flat. Because a wide range of surfaces are projected into a two-dimensional plane, the image must be distorted, so using a surface-scan camera is not effective, so you cannot use a surface-scan camera to fetch a full-dimensional surface at a time. Therefore, it is conceivable that the quality of the sealing element is checked in this way. It is rumored that the 4* person JL detection is more accurate' but it is not the most accurate detection result. To this end, the present invention provides an automated optical detection system and method thereof to improve the accuracy of the detection and improve the above-mentioned deficiency. [Summary of the Invention], the secret of this (4) is a kind of self-touching and sealing method. The surface area of the upper and lower surfaces of the element can be measured by the surface of the torus to improve the precision of the detection result. The other is to provide an automated optical inspection system and method thereof, the sealing component of the rpu#f, _ its non-transparent characteristics, and an infrared «light plate as a source of the moon, to image the sealing component of the PU material Improve the original _ glare can only see through the naked eye the shortcomings of the inner arc surface area. The clarification of the tree - the purpose is to provide an automated optical detection system and its method, which is to detect the inner and outer ring surface of the sealing element, the use The characteristics of the line scan camera, the image of the secondary capture-strip scan line, and the surface of the sample can be unfolded by the movement of the job, and the 2D image is processed for the image processing of the subsequent performance, thereby detecting (4) the component (10) torus The plague of the surface area. • This duck provides an automated optical inspection system and its method. The first-to-surface broom is used to take images and check the surface of the upper surface of the read component of the PU material. Detecting/seeing the surface area of the upper and lower surfaces of the sealing element, there is 瑕, 疵, that is, the detection is stopped. The surface area of the upper and lower surfaces of the sealing element is detected without flaws, and then the PU 麟 PU PU PU The component is imaged and detected, and the infrared duck (four) component is taken to be slightly transparent. The inner surface of the PU material is sealed, such as bubbles or foreign matter. The penetration of infrared rays is complete. The image of the arcuate area inside the sealing element is imaged. After the detection, it is found that there is flaw in the inside of the sealing element, that is, the detection is stopped. The inside of the sealing element is 200823448. After the sputum, the camera is scanned with a line of brooms. The surface area of the inner and outer annulus of the sealing element: ^ 瑕 仃 仃 仃 像 像 像 像 , 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测The trans-shipment of the surface m (four) two-dimensional image facilitates the image processing of the image, and the surface area of the inner surface of the sealing element is measured by the green; the surface area of the outer surface of the sealing element is found to be exhausted. (four) yuan The inside of the piece is still relied on, that is, the sealing element passed the test. The tree-shading device visual technology _ hair-set automatic optical inspection _ system, replacing the traditional twitching party ^. The test of the thief product causes an irreparable and serious loss when the fatigued seal is used in the expensive air hydraulic fluid machine. The following is a detailed explanation of the specific embodiment with the attached drawings. It is easier to understand the object, technical content, features and the effects achieved by the present invention. [Embodiment] The present invention provides an automated optical detection system and a method thereof, mainly provided in the first and second figures. A set of sealing element detection system comprising a surface detecting device, an internal lambda detecting device and a toroidal detecting device, thereby performing comprehensive flaw detection on the sealing member. The sealing member comprises a PU-packing, A product of a rubber seal (0il Seal), a gasket (Gasket) or a polyurethane seal (pu) material of any % seal. The detection of the sealing element includes the surface area of the upper and lower surfaces (shown in the first figure, the inner and outer ring surfaces (shown in the first figure, A), and the inner (the first figure, C) and other detection areas. Based on the sealing element (The object to be tested) is not a single plane. Therefore, the surface scan and the line scan camera are used to detect the sub-station according to the image capturing position. A total of 瑕8 200823448 can be detected. There are seven types including deformation, burrs, unclosed, and gas'. The object 'scratches, and foaming. The surface detection of step S10, the internal detection of the step-by-branch, the structure and details of the torus detection of the step side, and the reading slave of the present invention, __钱_细你 [Step S10 surface Detection] The detection process of the surface inspection of the step is as shown in the third figure, mainly detecting the surface of the upper and lower surfaces of the sealing element (Phase - Fig. B, B).
用一第-赌概賴轉，並透觀通舰縣械階二值化去除背景 等影像前處理；再進行步驟湖卿之辭徑變異數檢耻，最小平 方法將密封元件的内外徑邊界輪廓點，逼近出個別圓心與梅再進纖 所不，將個糊心與半徑與實際輪廓點至圓心的半徑距離作比較，利用統 計㈣變異量計算，以兩倍標準差作為上管制線、下管制線（既、既）。 如弟四圖所不’超出上管制線或制線的部分即為變形或毛邊瑕庇出現 處’如為超出上管制線或下管制線的部分，則進行步驟遞，檢測結束 並標示變形或毛邊位置。 若無變形或毛邊械出現’即無超出上管舰或下管制_部分，則 繼續進行未密合瑕疵檢測。 進行未密合瑕紐測時須先進行步驟S1G5、步驟及步驟8107，亦 即^標轉換、直柳化及投影_卜透過座標轉換，㈣環形的極 λ象轉換為4肖座；^的直條&像’為的是方便將未密合瑕i的特性(高 灰階對比)運用直方圖均值化與投影圖断，將此密封元件的瑕減更加突 來且使用直方圖均值化可將高對比的未密合瑕疫處灰階健高，再 200823448 ^用對比擴騎最大與最小灰離擴張至_55這兩娜限。接著使用投 ^也、向的像素累岭，便可根據判定是否為未密合瑕疲的標準，判斷密 、一疋u瑕龜’如觸結果為切合超過料範圍，貞彳進行步驟 _，檢·束並標示變形或毛邊位置；若未密合量在容許顏之内，則 進仃/驟卿之⑽檢測。換言之，在表面檢測裝置中檢測錢形、毛邊、 及未猜合这二種瑕軸壬_種，則此待聰财將被齡。反之，利用 内部檢測裝置繼續進行步驟S2G之内部檢測。Use a first-gamble to turn around, and through the Guanzhong County mechanical order binarization to remove the background and other image pre-processing; then proceed to the step of Hu Qing's rhyme variation, the minimum flat method will seal the inner and outer boundary of the component Contour points, approximating the individual center of the circle and the re-introduction of the plum, comparing the distance between the radius and the radius of the actual contour point to the center of the circle, using the statistical (four) variation calculation, using twice the standard deviation as the upper control line, Lower control line (both, both). If the part of the four figures does not exceed the upper control line or the part of the line is the deformation or the occurrence of the burrs. If it is beyond the upper or lower control line, the steps are handed, the test ends and the deformation or mark is marked. Burr position. If there is no deformation or burrs appearing 'that does not exceed the upper or lower control _ section, continue the non-closed 瑕疵 detection. Steps S1G5, steps and steps 8107 must be performed first, that is, the conversion of the standard mark, the straightening of the line, and the projection of the projections are performed by the coordinate conversion. (4) The polar λ image of the ring is converted into a 4th seat; Straight strip & like 'is convenient to use the histogram mean and projection map to break the characteristics of the unclosed 瑕i (high gray scale contrast), the reduction of the sealing element is more sudden and the histogram is averaged The high-contrast non-closed plague is gray-heeled, and then the 200823448 ^ expansion of the maximum and minimum gray-free expansion to _55 two limits. Then, using the pixel and the pixel, you can judge whether the secret, the 疋u瑕 turtle's result is more than the material range, and the step _, check • Beam and mark the position of the deformation or burr; if the amount of the occlusion is not within the allowable color, then the 仃/骤 (10) detection. In other words, in the surface detecting device, the money shape, the burrs, and the unspecified two kinds of axes are detected, and the money is to be aged. On the contrary, the internal detection of step S2G is continued by the internal detecting means.
【步驟S20之内部檢測】 山封元件在通過針對飾'毛邊、及未密合的檢測後，繼續使用内部 I懷置仏心封TL件^含錢泡或異物之械。密封元件内部形成氣 包或/、物之贼係由於射出成形的製程在注入縣狀態的即時，因壓力 或者溫度關題，使得㈣賴财之m法排域者錢物進入， 導致密封元制th現聽或異物辆種瑕疵。 内義测裝置’就疋將PU材質之密封元件内部（如第一圖所示，〇的 瑕錄測絲。由於pu材質_具透日職生，又因光線之齡波長愈長， 穿透性尤愈好。紅外線的波長，比任何一種可見光更長，所以穿透性 就更好因此本毛明使用紅外線背光板來打光，以將密封元件的内部之弧 面區域之喊突顯出來。本發明使用之紅外線背光板，如第五圖所示， 係使用182顆波長850奈米〜87〇奈米紅外線LED1〇，陣列式排放於1〇〇< 9〇麵尺寸義包板11上，於紅外線上方設置擴舰9⑽fuser)， 俾使發出光線能夠均勻。 200823448 第一面掃瞒攝影機7,係採用CCD感光元件來實現圖像採集。而c⑶感 光讀有-個特性：它的感光範圍在波長咖奈米〜·奈米之間；換言 之絲树關_朗人騎能·的可見光線與部分人眼不可見 的紅外光線。面掃鳴影機用鏡頭8，係使用可變焦距鏡頭_一ι〇χ㈣ 爾，以便對不同尺寸的密封元件進行取像。 以上"紹本發明使m狀理&，以下晚續說測的檢測流 程，如第六圖所示’進行步驟s謝，利用紅外線背光板打光取像，然後繼 • 續進行S202歸驟湖，利用低通渡波及平均灰階二值化去除背景等影像 前處理及座標轉換；接續進行步驟S204及S205，使用正規化灰階變異數檢 測法，以進行瑕疲的分類篩檢，其為一種針對㈣封元件影像的演算法， 如通過正規化灰階變異數檢測法之檢測，未檢_氣泡或異物，即繼續進 行步驟S30之環面檢測。 如未通過正規化灰階變異數檢測法之檢測，則繼續進行步驟s細及 S207，使用物件分析的方法（B1〇b分析），繼續精確的_是否真__ 馨 I物之瑕症存在。 如判斷真有氣泡與異物之瑕龜存在則進行步驟S2〇8及§删，結束檢 測’此密封元件係一瑕疵品。 步驟S206檢測後之結果，並無瑕疲；則進入步驟咖之環面檢測繼 續檢查其他的瑕疵種類，如刮傷或發泡。 上述之紅外線背光板需均勻排列，本發明所使用的紅外線⑽共啦 顆’使用可調式電源供應器，電壓由3Volt〜雨。lw_欲檢測㈣元件2 200823448 樣本厚度作調整，厚度5顧有效檢測亮度約為3· 7 Volt〜4.4Volt ;厚度 10肅有效檢測亮度4. 3 Volt〜5. 7Volt。 【步驟S30之環面檢測】 步驟S30為環面檢測，主要目的是檢查有沒有刮傷或發泡這兩種瑕症， 出現在岔封元件的内外環表面（第一圖所示，A)。第三檢測站係採用線掃瞒 攝影機對迫緊之内外環表面之表面區域進行取像外，檢測方法與内部檢測 之方式相同。 再來說明環面檢測裝置，其架構可參照如第七圖所示，將兩台線掃描 攝影機1、2分別架設如第七圖所示，線掃描攝影機卜2上分別·具有線掃 描攝影機用鏡頭3，其中兩台攝影機的夾角約10度〜2〇度，將待測之密封 元件5固定於轉盤6上（須注意偏心問題）。 線掃描攝影機取像頻率（Hz)為垂直解析度除以取像時間（sec)的關 係，若欲在0· 18秒取完影像，馬達4轉速需333 rpm，便可即時完成取像。 壞面檢測流程可參照如第八圖所示，進行步驟S3〇1及步驟兑〇2，利用 線掃描攝影機對密航件之内外環面之表面區域進行取像，並利用低通遽 波及平均灰階二值化等影像前處理去除背景。由於線掃描取出的影像為長 條形，因此線掃描攝影機取像後，毋需進行座標轉換；接續進行步驟s3〇3 及S304，使用正規化灰階變異數檢測法，以進行瑕疵的分轉檢，如通過 正規化灰_舰制奴_，級_觸紐泡，即為步驟谓之 通過二道檢測手續之密封元件。 如未通過正規化灰階Μ數㈣法之—，_軌行步驟通及 12 200823448 S206使用物件刀析的方法⑻此分析)，繼續精確的判斷是否真有刮傷或 發泡瑕疵存在。 如步驟S3G6判斷之結果，並無瑕u即為步驟s腳之通過三道檢測 +、A㈣元件。如步驟S3G6判斷之結果，有刮傷或發泡瑕窥;則如步驟 S308及S309，結束檢測，此密封元件係一瑕龜品，有刮傷或發泡瑕疲。 另外，由於密封元件的脣形部位稍有傾#(如第一圖所示，D)，因此 使用線掃描攝影機錄外環表面時，f將掃描線與取像表面平行；所以需 籲稍,微傾斜、力5度。值得一提的是，擷取内環表面時會擷取到不必要的上表 面⑻，因此必需透過步驟·之影像前處理更改檢測範圍，兩線掃描攝 影機間隔15度時，是較佳的選擇。再者，使用線型光纖光源打光，可使光 源集中於㈣fe線上，打光角度為與擁取面的法線夾角約2〇〜洲度，光源亮 度5000〜10000勒克斯(lux)。 藉由上述清楚之制，本發明係針對pu#f之密封元件，可應用之範 瓣射出成型業、化卫業、...等。本發明係—套以機器視覺為基礎 #㈣動化光學檢測系統，尤指—種使用線掃描攝影機對環形曲面進行取像 的方式，以及專為PU材質㈣元件所設計的制紅外料光紐樣本之打 . 綠式’可取代傳統人工檢測密封元件的方式。狀密封元件的形狀與材 f的雜，本發_職狀件的雜結構，使魏贼轉運祕配線掃 描攝影機，達成三度空間曲面成為二維影像的可能性。另外，運用pu材質 本身稍具透明.的特性，採用紅外線LED所組成的背光板，目的是將猶具透 明特性的晴質密封元件之内部之弧面區域之祕，譬如氣泡或是異物， 13 200823448 透過紅外線的穿透力完整地將瑕疵成像出來。並以後續的電腦影像處理作 為瑕疫分析之手段，取代傳統利用肉眼於強光下之環境，内部若有瑕龜產 生，會使光線在密封元件内部形成多次折射的因素來檢測，本發明藉此改 進此一既花成本又不理想的檢測方式。 經實際测試後發現，本發明之自動光學檢測系統，其檢測時間會因瑕 疵在哪一站被檢測出來而有所不同。檢測一個樣本的時間最快為〇 7耵和、。 對無瑕疵之樣本來說，由於必須通過三次的檢測，所花費的時間最長，檢 查-個需5.4秒。瑕紐出率為90· 1%;有瑕疲而未被發現者有⑴固，因 此漏測率為5· 1% ;無贼而被誤·贼品者有10個，因此假警報率為 4· 7%。 …、 惟以上所述者，僅為本發明之較佳實施例而已，並非用來限定本發明 實施之細。故凡依本發财請細所述之形狀、構造、特徵及精抑為 之均等*化或修飾，均應包括於本發明之巾請專利範圍内。 ”、、 【圖式簡單說明】 第一圖係本發明之密封元件示意圖。 第二圖係本發明之檢測密封元件之檢測流程示意圖。 第二圖係本發明之表面檢測之流程示意圖。 第四圖係本發明判斷變形與毛邊瑕疵之數據圖。 第五圖係本發明之内部檢測裝置之架構圖。 第六圖係本發明之内部檢測之流程示意圖。 第七圖係本發明之環面檢測裝置之架構圖。 200823448 第八圖係本發明之環面檢測之流程示意圖。 【主要元件符號說明】 - 1線掃描攝影機 2線掃描攝影機 3線掃描攝影機用鏡頭 4馬達 5密封元件 ^ 6轉盤 7第二面掃瞄攝影機 8面掃瞄攝影機用鏡頭 9擴散膜 10紅外線LED 11麵包板 15[Internal Detection in Step S20] After the detection of the trimmed trim and the non-closed seal, the seal element continues to use the internal I to seal the TL piece ^ containing the money bubble or the foreign object. The thief system that forms the air bag or/or the inside of the sealing element is in the state of injection into the county immediately due to the pressure or temperature problem, so that the money is entered by the m method, which leads to the sealing element system. Th is listening or foreign objects. The internal measuring device 'is inside the sealing element of the PU material (as shown in the first figure, the 测 瑕 测 测 。 。 。 。 由于 由于 由于 由于 由于 由于 由于 由于 由于 由于 由于 由于 由于 由于 由于 由于 由于 由于 由于 由于 由于 由于 由于 由于 由于 由于 由于 由于The better the sex is. The wavelength of the infrared light is longer than any kind of visible light, so the penetration is better. Therefore, the hair is illuminated by an infrared backlight to highlight the arc of the inner surface of the sealing element. The infrared backlight board used in the present invention, as shown in the fifth figure, uses 182 wavelengths of 850 nm to 87 Å nanometer infrared LED 1 〇, and the array is discharged on a 1 〇〇 < 9 〇 尺寸 义 义 板 11 , set the expansion ship 9 (10) fuser above the infrared ray, so that the light can be even. 200823448 The first broom camera 7 uses CCD sensor to achieve image acquisition. The c(3) sensible reading has a characteristic: its sensitization range is between the wavelengths of the nanometers and the nanometers; in other words, the silk trees are closed to the visible light rays of the people and the infrared rays of some human eyes are invisible. The lens 8 for the face sweeper uses a zoom lens _ ι 〇χ (4) to capture images of different sizes of sealing components. The above "Shao this invention makes m-like & the following test flow of the test, as shown in the sixth figure, 'steps s thanks, using the infrared backlight to illuminate the image, and then continue S202 return In the lake, the low-pass wave and the average gray-scale binarization are used to remove the background image processing and coordinate conversion; the steps S204 and S205 are successively performed, and the normalized gray-scale variation number detection method is used to perform the classification screening of the fatigue. It is an algorithm for (4) sealing the image of the component. For example, by detecting the normalized gray-scale variation number detection method, the undetected bubble or foreign matter, that is, the torus detection of step S30 is continued. If the detection of the normalized gray-scale variability test is not passed, proceed to step s and S207, and use the method of object analysis (B1〇b analysis) to continue the accurate _ whether or not __ . If it is judged that there is a bubble and a foreign body, the step S2〇8 and § are deleted, and the detection is terminated. After the result of the test in step S206, there is no fatigue; then the torus test of the step coffee is continued to check other types of defects such as scratches or foaming. The above-mentioned infrared backlight panels need to be evenly arranged, and the infrared rays (10) used in the present invention use an adjustable power supply, and the voltage is from 3 Volt to rain. Lv_ want to detect (4) component 2 200823448 sample thickness for adjustment, the thickness of 5 effective detection brightness is about 3.7 Volt~4.4Volt; thickness 10 effective detection brightness 4. 3 Volt~5. 7Volt. [Hollow surface detection of step S30] Step S30 is toroidal detection, and the main purpose is to check whether there are two types of hysteria such as scratching or foaming, which appear on the inner and outer ring surfaces of the sealing element (shown in the first figure, A) . The third detection station uses a line broom camera to take images of the surface area of the inner and outer rings that are tight, and the detection method is the same as the internal detection. Next, the toroidal detecting device will be described. The structure can be referred to as shown in the seventh figure. The two line scanning cameras 1 and 2 are respectively arranged as shown in the seventh figure, and the line scanning cameras are respectively provided with the line scanning camera. The lens 3, in which the two cameras have an angle of about 10 degrees to 2 degrees, fixes the sealing member 5 to be tested on the turntable 6 (note the eccentricity problem). The line scan camera's image capture frequency (Hz) is the vertical resolution divided by the image capture time (sec). If you want to take the image at 0. 18 seconds, the motor 4 speed needs 333 rpm, and the image can be taken instantly. For the bad surface detection process, refer to step S3〇1 and step 〇2 as shown in the eighth figure. The line scan camera is used to image the surface area of the inner and outer annulus of the escrow piece, and the low-pass chops and the average ash are utilized. Image preprocessing such as order binarization removes the background. Since the image taken by the line scan is a long strip shape, after the line scan camera takes the image, coordinate conversion is not required; the steps s3〇3 and S304 are successively performed, and the normalized gray-scale variation number detection method is used to perform the splitting of the 瑕疵. Inspection, such as through the formalization of ash _ ship slave _, level _ touch bubble, that is, the step is said to pass the two inspection procedures of the sealing components. If the normalized gray scale number (four) method is not adopted, the _ rail step is passed and 12 200823448 S206 uses the method of object analysis (8) for this analysis, and it is necessary to continue to accurately determine whether there is a scratch or foaming flaw. As a result of the judgment in the step S3G6, there is no 瑕u, that is, the three-pass detection +, A (four) elements are passed through the step s. If the result of the judgment in the step S3G6 is that there is scratching or foaming, the detection is ended in steps S308 and S309, and the sealing member is a turtle, which is scratched or foamed. In addition, since the lip portion of the sealing member is slightly inclined # (as shown in the first figure, D), when the surface of the outer ring is recorded by the line scanning camera, f scans the scanning line parallel to the image capturing surface; Micro tilt, force 5 degrees. It is worth mentioning that when the inner ring surface is captured, an unnecessary upper surface (8) is captured. Therefore, it is necessary to change the detection range through the image pre-processing of the step, and the two-line scanning camera is a better choice when it is separated by 15 degrees. . Furthermore, using a linear fiber optic source to illuminate the light source can be concentrated on the (four)fe line, the angle of illumination is about 2〇~ continents from the normal of the gripping surface, and the light source brightness is 5000~10000 lux. With the above clear system, the present invention is directed to the sealing member of pu#f, which can be applied to the molding industry, the chemical industry, and the like. The invention is based on machine vision. (4) Dynamic optical detection system, especially a method for taking a circular curved surface using a line scan camera, and an infrared material light button designed for a PU material (four) component. The sample type. The green type can replace the traditional way of manually detecting the sealing element. The shape of the sealing element and the miscellaneous structure of the material f, the hybrid structure of the hair _ job, so that the Wei thief transports the wiring to scan the camera, achieving the possibility that the three-dimensional surface becomes a two-dimensional image. In addition, the pu material itself is slightly transparent, and the backlight board composed of infrared LEDs is used to make the secret of the inner arc surface of the transparent sealing element which is transparent, such as bubbles or foreign objects. 200823448 Completely image the cockroach through the penetration of infrared light. And the subsequent computer image processing is used as a means of plague analysis, instead of the traditional use of the naked eye under the strong light environment, if there is a turtle inside, the light will be detected within the sealing element to form multiple refraction factors, the present invention This improves the detection method that is both costly and unsatisfactory. It has been found through actual testing that the detection time of the automatic optical inspection system of the present invention varies depending on which station is detected. The fastest time to test a sample is 耵 7耵 and . For innocent samples, it takes the longest time to pass three tests, and it takes 5.4 seconds to check. The rate of 瑕 瑕 90 is 9·1%; those who are unsatisfied and undetected have (1) solid, so the rate of missed detection is 5.1%; 4.7%. The above is only the preferred embodiment of the present invention and is not intended to limit the implementation of the present invention. Therefore, the shape, structure, characteristics and refinement of the above-mentioned wealth, please be equal to the scope of the patent application of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS The first drawing is a schematic view of the sealing element of the present invention. The second drawing is a schematic diagram of the detection process of the detecting sealing element of the present invention. The second drawing is a schematic flow chart of the surface inspection of the present invention. The figure is a data diagram of the internal deformation detecting device of the present invention. The fifth drawing is a schematic diagram of the internal detection device of the present invention. The seventh drawing is a schematic flow chart of the internal detection of the present invention. The structure diagram of the device. 200823448 The eighth diagram is the flow chart of the toroidal inspection of the present invention. [Main component symbol description] - 1-line scanning camera 2-line scanning camera 3-line scanning camera lens 4 motor 5 sealing element ^ 6 turntable 7 Second side scanning camera 8 side scanning camera lens 9 diffusion film 10 infrared LED 11 bread board 15