200905577 九、發明說明： * 【發明所屬之技術領域】 本發明涉及一種虹膜辨識系統，尤指一種可增加辨識正確性和資料處 理效率之虹膜辨識系統。 【先前技術】 隨著社會經濟的發展，人們對於安全的問題越來越重視，生物辨識為 近年來非常糾的研究題目。目前在生物_領域中，行為特徵作為 辨識方法巾較常見的為透過聲音⑽ee)、簽名(signatoe)進行辨識；而採用 生理特徵的方法财臉师aee)、減(fm卿int)、視_(__、_ (iris)、掌紋(palmprint)以及掌型(hand)等等。生物辨識即利用身體本身特有 的特徵來做為識別體。由於人類有許多生物特徵是獨—無二，加上這此特 徵疋跟隨本人，不用擔心被有心人士竊取。200905577 IX. INSTRUCTIONS: * Technical Field of the Invention The present invention relates to an iris recognition system, and more particularly to an iris recognition system capable of increasing recognition accuracy and data processing efficiency. [Prior Art] With the development of social economy, people pay more and more attention to the issue of safety. Biometric identification is a research topic that has been very difficult in recent years. At present, in the biological field, the behavioral feature is more commonly recognized as the identification method towel through the sound (10) ee) and the signature (signatoe); while the physiological feature method is used by the financial face aee), minus (fm qing int), visual _ (__, _ (iris), palmprint (palmprint) and palm (hand), etc. Biometrics use the unique characteristics of the body as a recognition body. Because humans have many biological characteristics are unique - no, plus This feature is to follow me, don't worry about being stolen by someone who is interested.
人最普遍為人所知的生物辨識就是指紋辨識，它有著高度的方便性與安 王性’不需要記住複雜的密碼，也不需隨錢帶餘匙、智能卡之類的東西。 然而’指紋在過去的研究中，還是容易受到外在因素改變而降低其辨識率 的缺點’因此日後就有視網膜掃描辨識技術被提出。視嗔掃描一产成為 身份辨識的主料具之―，但是視網膜影像取得必須透過紅外線翻，而 T或蝴酬人眼恤械繼，因此，新的虹膜辨識技術 /之而生。自198^Le〇nardF1⑽―兩位美籍的眼科醫生， 率先虹膜的特徵作為生物辨識的依據之後，隨著人們對於資 要求提升’虹膜_的高觸紅漸漸在市場上縣_，重要性已非同 5 200905577 日而語 域是指瞳孔觸有顏色的肌_織，人的虹膜上魏多微小的凹凸 起伏和條狀時具娜輯。.人_其—纽戦幾乎不會 產生任何變化，且其外部包裹著透明嫩，不易遭受外力的傷害而改變。 -有箸嶺肖卿爾，爛朗她偷何部分都多， 例如膠梅、赚、«、色素、蛇_束、橫晴所组成… 共有二百时咖㈣，做τ，卿心彻爾，指紋只有 —十到四十個。據研究指出，兩個人的虹媒相同的或然率為妒分之一， 即使是同樣—個人，左右兩_虹膜也有著各自的結構。 的過程與指紋識別類似，需將掃描的虹膜圖像轉換為數位代 碼，存儲到電腦資料庫。當進行身 僳昭s _ 細私如辦，只需輯待檢測者的虹膜圖 像照，即可判明雜，湘 術中最#。 闕_«較各觀物辨識技The most commonly known biometric identification is fingerprint recognition. It has a high degree of convenience and security. It does not need to remember complicated passwords, and does not need to carry extras, smart cards and the like. However, in the past research, fingerprints are still vulnerable to external factors and their recognition rate is reduced. Therefore, retinal scanning recognition technology has been proposed in the future. Vision scanning has become the main tool for identification, but the retina image acquisition must be turned through the infrared rays, and the T or the rewards are followed by the eye, so the new iris recognition technology is born. Since 198^Le〇nardF1(10), two American ophthalmologists, took the lead in the characteristics of the iris as the basis for biometric identification, and as people's demand for the increase of 'Iris' is becoming more and more popular in the market, the importance has been Not the same as 200905577 and the language field refers to the pupil of the color of the muscle _ woven, the human iris on the Wei multi-fine undulations and strips when the Na series. People _ their 戦 戦 戦 戦 戦 戦 戦 戦 戦 戦 戦 戦 戦 戦 戦 戦 戦 戦 戦 戦 戦 戦 戦 戦 戦 戦 戦 戦 戦- There is Yuling Xiaoqinger, and she has to steal more parts, such as Jiaomei, earning, «, pigment, snake_bundle, and Hengqing. There are two hundred coffees (four), do τ, Qing Xinchel The fingerprints are only - ten to forty. According to the research, the same probability of the two people's rainbow media is one of the points, even the same - individual, the left and right _ iris also have their own structure. The process is similar to fingerprint recognition. The scanned iris image needs to be converted to digital code and stored in a computer database. When you carry out the body 僳 s _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _阙_«Compared with each other
【發明内容】 ^曰加辨識正雜和資料處觀料虹膜辨 本發明之目的係提供一種 識糸統。 其包含一影像擷取裝置、 “本發明之_實施例提供—種虹顧識系統， 輪型資料庫、—影像前置處職組 模組。，文抽取處理模組以及一特徵比對 t且轉置係絲練 用來儲存财㈣ 者之眼睛影像。該模型資料庫係 謂存複數個虹膜影像資料 办像則置處理模組包含—瞳單 6 200905577 元:一虹膜定位單元、-域影像正規化單元以及—影像增強單元。該瞳 孔定位單元用來依據該眼睛影蚊位該眼睛影像之瞳孔。該虹膜定位單= 用來以該眼睛影像之睛孔之圓心像素為圓心，選取一第一半徑為圓周之: 第-像素組以及-第二半徑為關之—第二像麵，並依據該第_像素組 之總和以及該第二像素組之總和之差之絕對值，決定該眼睛影像之虹膜區 域。該虹膜影像正規化單元用來正規化該虹臈區域以輸出—正規化虹膜區 域。騎像增強單元用來等化該正規化虹膜區域以產生一等化虹膜 該特徵抽取處理模組係用來抽取該等化虹膜區域之特徵。該特徵比對模組 係用來比對該等化虹廳_的概以及魏個虹歸彡像資料。 依據本發明之-實施例，該瞳孔定位單元用來二值化該眼睛影像，並 自該二值化眼睛影像上選取—第三半徑為圓周之—第三像素組以及一第四 半徑為圓周之-第四像素組，並依據該第三像素組之總和以及該第四像素 組之總和之差之絕對值，定雌轉f彡像之瞳孔。 依據本發明之—實施例，該影像擷取裝置包含一第一攝影裝置、—判 斷單元、—第屬彡裝⑽—調嶋„裝蝴來拍攝該 使用者之臉像。該觸單域时觸雜騎之臉«像之眼睛部 份是否織咖㈣酬，高健舞糊斷單元 判斷該使用者之臉部影像之眼睛部份位於該臉部影像之預設位置時，攝取 該使用者之眼崎。瓣_梅爛_雜使用者之臉 部影像之眼睛赌並摊於該臉部影像之職健時，調_第二攝影裝 置之拍攝位置。其中該調整機構包含—步進馬逹。該影像擷取裝置另包含 200905577 一紅外線投射器，用來射出一波長範圍為7〇〇_9〇〇nm之紅外線，且該第— 攝影裝置以及雜二攝織置冑包含_紅外線麟(IR Fil㈣，絲使得只 有波長範ϋ在7GG__nm之紅外線通過。該觸單元伽來二值化該臉部 衫像，並自该二值化臉部眼睛影像上選取一第五半徑為圓周之一第五像素 組以及第，、半控為圓周之一第六像素組，並依據該第五像素組之總和以 及該第六像綠之總和之差之絕驗，判_個者之臉部影像之眼睛部 份是否位於該臉部影像之預設位置。 依據本發明之-實施例’該虹膜影像正規化單元制來將該虹膜區域自 極座標轉換為垂直座標。 依據本發日狀—實關，該舰輯模組雜収㈣量機卜伽 machine)。 .依據本發卜該細A取處賴組铜來則、波轉換的方式 抽取該等化虹膜區域之特徵。 為讓本發明之上述和其他目的、特徵、和優點能更鴨紐，配合所 附圖式，作詳細說明如下： 【實施方式】 請參閱第1圖，第1圖係本發明之虹膜辨識系統1〇之功能方塊圖。虹 膜辨識系統10包含一影像擷取裝置5〇、—旦 心像則置處理模組2〇、一特徵 抽取處理模組30以及一特徵比對模組4〇 汉模型貧料庫50。影像擷取 200905577 係用來拍攝使用者之目_像5。影像前置處理模組洲來於接收 一眼睛鱗5之後，對鴨影像5進行前置處理辦，以去除縣、瞳孔 部份’並將虹膜影像自眼睛影像5取出。特徵抽取處理模組3Q將虹膜影像 的特徵抽取出來。虹膜辨識系統1〇可以分為使用者註冊以及辨識使用者兩 個流程。註冊個者時，姆攝使用者人眼影像數張，在影像前置處理 板組2以及特徵抽取處理輸3G處理後，會將仙者域影像資料存入 模型資料庫50中。觸制者時，影像同樣先經過影像前置處理模組如 以及特徵抽取處理模組3〇處理後，再利用特徵比對模組4〇比對模型資料 庫50的内容以產生結果。 請參閱第2圖，第2圖係第1圖之影像梅《置誕功能方塊圖。影 像擷取裝置50包含-第，跑52、L彡健54、—判斷單元 56、-調整機構58以及—紅外線投射器55。紅外線投_5絲射出波 長範圍為•nm之紅外線。祕其波長雜人_狀—種不可見光， 即使對人眼近距離騎也不會有不適感，加上紅外光對於物體也有良好的 反射作用，可以將虹膜_紋理清晰的反射出來。第—攝影裝置52以及第 二攝影裝置54之鏡猶，另包含—紅外線麵則㈣，只讓彻塌· 的紅光線通過，濾除掉其他可絲對影鮮來的影響。為了達到大範圍之 自動追蹤效果’本實施例之影像擷取裝置5〇使用第一攝影裝置Μ，其具有 視野較大之人麟影綱細取制者之臉娜像，綱料—攝影裝置 52所擷取之影像’搭配人臉横測之機余j，將影像搁取裝置％做第一階段的 定位。此階段的定位可將人眼影像大略定位於人眼取像的鏡頭，如果畫面 200905577 ㈣出現人眼的影像’此第—攝影裝置52將會由中心向外做小範圍的搜 尋。第1段_人眼取像的第二攝影裝置54來做準確的定位。第 中，在第：攝影裝置54每—步移動之後，判斷單元兄必須判斷晝面中^ 否有鳩細辦嘴，讀抓目晴彌動作。疋 請一併參閱第2圖以及第3圖，第3圖係第】圖之影像搁取袭置 取眼睛影像之流程圖。其包含下列步驟： 掏 步驟300 :第-攝影裝置52拍攝—使用者之臉部影像。 驟310 步驟露判斷該臉部影像之眼睛部份是否位於該臉部影像之預設位置 本實施例中，判斷^56會判斷該臉部影像之眼睛部份是 於该臉部f彡像之中心位置。若是，執行步_，若否，執行斗 步驟3〇4 ’第—攝影裝置Μ拍攝使用者之眼睛影像。 步驟306 :判斷單元56判斷眼睛影像是否位於畫面中心，” 驟 兕8,若否，執行步驟312。 右疋仃步 步驟:舰眼睛影像《彡像前置處理單元2〇。 步驟仙：調整物移動調整第,裝置52之拍攝位置 58利用—步進馬達，故可微調第-攝影”52攝位置構 步驟312 ·調細構58飾聊：嶋置Μ㈣位置，調 58係利用—步進馬達，故可微二攝影裝置52之拍攝位置 200905577 第-攝影裝置52拍攝臉部影像(步驟3〇〇)之後，判斷單元&會判斷眼 睛部份是否位於臉部影像之中心位置。由於眼睛之瞳孔為一個灰階值極 低，呈現圓形的區域，且位於人_中心，因此判斷單元52將利用瞳孔的 存在與否’作為綱眼睛部份驗置。首先機單元Μ會與帛—攝影裝置 52拍攝臉部影像之所有像素輯—臨界灰階值做比較，當像雜之灰階值 大於該臨界讀辦’減像素狀灰·設定為255，狀，當像素點之 灰階值小於該臨界灰階值時，把該像素點之灰階值設定為卜如此一來，臉 部影像會呈現一個像素二值化影像。 凊-併參M 4A_4C ® 1 4A_4D目侧斷瞳孔半徑細之示意圖。 第攝衫裝置52所拍攝的臉部影像將過二值化處理後，會呈現如第仏圖 所示之影像。因為曈孔具有近似圓形的雜，以下將彻―種稱之為測圓 機制來决疋瞳孔的位置。因為瞳孔在人眼影像中為像素值最低的區域，所 以瞳孔的邊界就是存在於灰階值變化最大的地方。也就是說，計算内圓(半 徑m-i)圓周上各點之像素值總合與外圓(半徑r〇=r+1)圓周上各點之像素值 總合之差，若出現差值最大的圓形區域即判斷該圓型區域為瞳孔的位置。 舉例來說’判斷單元56會建立一個以像素點〇為圓心、，兩個内外徑分別為 I '及r〇之同、圓周遮罩，其中Γι<1〇 , r〇>3〇。在η的圓周上取1〇個像素 點A1 Α2、·..、Α10 ’在r〇的圓周上亦取1〇個像素點Β1、Β2、...、Β1〇， 則内圓周上邊點像素值尸/=jP⑽+ Ρ⑽+ · + ρ(·)以及外圓周上邊點像素 值總和為/W㈣+ _) + ··. + />_。並計算外圓以及内圓像素總合之差 Ρ = 卜對於在大小範圍之内的曈孔，會得到最高的差值(如第4(：圖所 200905577 不)。當轉元5_具有最高的朗，即職部断含有眼睛部 份區域，且判斷單元會判斷像素點〇是否位於臉部影像之中心位置。如果 f攝影裝置Μ拍攝時，使用者的臉部有所偏差而未拍攝到眼睛，或是眼 睛部份亚未位於臉部影像中間位置，則調整機構％會調整第—攝影裝置^ Z 之臉部影像，直到判斷單元Μ判斷臉部影像 t ^有眼目，且_部份位於臉娜像巾間位置。 一旦判斷單元56 __像中含有_域，且_ 〇位於 _像_ _，二_置54會_—攝繼52所決 定出來的定位點拍攝使用者眼 u 月如像。較佳地，第二攝影裝置54的像素數 攝影裝置52多’且其影像品質較佳，所以第二攝影裝置54會強 攝的疋解析度較佳的眼睛影像。接下來，判崎置％會湘上述的測 圓機制判斷第二攝影54 + •^置54操取出來的眼睛影像的瞳孔是否位於中心位 置。若瞳孔位於該眼睛影像的中心 彳4靡忒眼睛影像5傳送至影像前 早凡〇8作下—步的處理；^否，則調整機構312會調整第二攝影 装置54的拍攝位置，織再次拍攝使用者之眼睛影像，直到判斷單元56 判斷瞳孔位於眼睛影像中間位置。 i閱第5圖’第5圖係第丨圖之影像前置處理模組如之功能方塊圖。 影像前置處理她2Q包含—曈孔定位料22、—虹歡位單元^、一虹 膜影像正航單元26以及一影像強化單元Μ。當眼睛影像$被擁取出來 後，影像前置處理模㈣細曈孔定位單元22對眼睛影像$的瞳孔定 位以決定眼睛影像之圓心。接私觀位單元Μ依據目崎影像之圓心對眼 12 200905577 睛影像5之虹膜邊界做定位，標示出虹膜 所在之環狀區域。之後利用虹膜 影像正規化單元26對虹膜 内位以及物做正規化的動作，將躲虹膜影像 縮放至具有相_内外徑大小。為了提高辨識的效果， 等化影像的方式來加強虹臈内的紋理資訊。 影像強化單元28會 由於輸入的眼睛影像5只要求包含虫工膜以及瞳孔部分的影像，瞳孔的 位置不-定在眼睛影像5的中心、。^虹膜係環繞於瞳孔之圓型物，所以 要判斷虹膜的位置，必須先決定瞳孔驗置。由於軌灰階值偏低的特性， 瞳孔疋位早π 22會先將_影像5之所有像総與—臨界灰階值做比較， 當像素點之灰階值大於該臨界灰階值時，把該像素點之灰階值設定為况， 反之，當像素狀灰_小_臨界統_，把輯素點之灰階值設定 心如此-來，眼睛影像會呈現—個像素二值化眼睛影像。因躺孔半徑 祀圍4在10-30像素大小，且瞳孔在人眼影像巾為像素值最低的區域，所 以瞳孔的邊界就是存在於灰階值變化大的地方。也就是說，計算峨半徑 )圓周上各點之像素值總合與外圓(半徑听+1)圓肖上各點之像素值總 口之差’若出現差録大辭徑f(㈣）/2的_區域㈣_圓型區域 為瞳孔的位置。舉例來說，瞳孔定位單元22會建立-個以像素點q為圓心， 兩個内外分別為Π以及之同心圓周遮罩，其中抑，明〇。在^的圓 〇 個像素點A1、A2 '…、A10，在r〇的圓周上亦取1〇個像素點SUMMARY OF THE INVENTION The object of the present invention is to provide a system for identifying the nucleus. The utility model comprises an image capturing device, “the invention provides a seeding system, a wheel type database, an image front working group module, a text extraction processing module and a feature comparison t And the transposition silk is used to store the images of the eyes of the person (four). The model database is a plurality of iris image data, and the processing module is included - 瞳 单 6 200905577 yuan: an iris positioning unit, - domain The image normalization unit and the image enhancement unit are configured to: according to the pupil of the eye image, the iris positioning unit is used to select a center of the center pixel of the eye hole of the eye image. The first radius is a circle: the first pixel group and the second radius are the second image plane, and the absolute value of the difference between the sum of the _ pixel groups and the sum of the second pixel groups is determined. An iris region of the eye image. The iris image normalization unit is configured to normalize the rainbow trout region to output a normalized iris region. The riding image enhancement unit is configured to equalize the normalized iris region to generate an equalized iris. The feature extraction processing module is configured to extract features of the equalized iris region, and the feature comparison module is used to compare the image of the rainbow hall and the Wei Gehong image. According to the present invention - In an embodiment, the pupil positioning unit is configured to binarize the eye image, and select from the binarized eye image—the third radius is a circle—a third pixel group and a fourth radius is a circumference-fourth pixel Grouping, and determining the pupil of the image according to the sum of the sum of the third pixel group and the sum of the fourth pixel groups. According to an embodiment of the invention, the image capturing device comprises a first A photographic device, a judging unit, a singular armor (10), a singer, is used to photograph the face of the user. When the touch field is touched, the face of the rider is photographed. If the eye part of the user's face image is at the preset position of the face image, the eye part of the face image of the user is determined. Ingest the eye of the user.瓣_梅烂_ The face of the user's face When the eye of the image is gambled and spread on the face of the face image, adjust the shooting position of the second camera. Wherein the adjustment mechanism comprises a stepping horse. The image capturing device further comprises a 200905577-infrared projector for emitting infrared rays having a wavelength range of 7 〇〇 _9 〇〇 nm, and the first photographic device and the hybrid photographic device comprise _ infrared ray (IR Fil (4), the wire allows only the wavelength range to pass through the infrared rays of 7GG__nm. The touch unit gamma binarizes the face shirt image, and selects a fifth radius from the binarized facial eye image as a fifth of the circumference The pixel group and the first, the half control is a sixth pixel group of the circumference, and according to the sum of the sum of the fifth pixel group and the sum of the sixth image greens, the eyes of the face image of the individual are judged Whether the portion is located at a preset position of the facial image. According to the embodiment of the present invention, the iris image normalization unit is configured to convert the iris region from a polar coordinate to a vertical coordinate. According to the present invention, the Shipbuilding module miscellaneous (four) measuring machine Bu Jia machine). According to the hair of the present invention, the characteristics of the equalized iris region are extracted by means of the copper group and the wave conversion. The above and other objects, features, and advantages of the present invention will be described in detail with reference to the accompanying drawings. FIG. 1 and FIG. 1 are the iris identification system of the present invention. 1〇 function block diagram. The iris recognition system 10 includes an image capture device 5, a cardiac image processing module 2, a feature extraction processing module 30, and a feature comparison module 4. Image capture 200905577 is used to capture the user's eye _ image 5. After the image pre-processing module receives the eye scale 5, the duck image 5 is pre-processed to remove the county and pupil portions and the iris image is taken out from the eye image 5. The feature extraction processing module 3Q extracts the features of the iris image. The iris recognition system 1 can be divided into two processes: user registration and identification of users. When registering an individual, a number of human eye images are captured, and the image data of the immortal domain is stored in the model database 50 after the image pre-processing panel group 2 and the feature extraction processing 3G processing. In the case of the toucher, the image is first processed by the image pre-processing module and the feature extraction processing module 3, and then the feature comparison module 4 is used to compare the contents of the model database 50 to produce a result. Please refer to Figure 2, which is the block diagram of the image of the image of Figure 1. The image capturing device 50 includes -, RUN 52, L 彡 54, - determination unit 56, - adjustment mechanism 58, and - infrared projector 55. Infrared projection _5 wire emits infrared light with a wavelength range of • nm. The secret wavelength of the _ _ _ invisible light, even if the human eye is not close to riding, and the infrared light has a good reflection effect on the object, the iris _ texture can be clearly reflected. The first photographing device 52 and the mirror of the second photographing device 54 further include an infrared surface (four), which allows only the red light to pass through, and filters out the influence of other filaments on the image. In order to achieve a wide range of automatic tracking effects, the image capturing device 5 of the present embodiment uses a first photographic device Μ, which has a larger field of view, and is a face-photographing device. 52 captured images 'matching the face of the face measurement machine j, the image capture device % to do the first stage of positioning. The positioning at this stage can roughly locate the human eye image in the lens of the human eye. If the image of the human eye appears on the screen 200905577 (fourth), the camera device 52 will perform a small search from the center outward. The first segment _ the second imaging device 54 of the human eye image is used for accurate positioning. In the middle, after the photographing device 54 moves every step, it is judged that the unit brother must judge whether there is a fine mouth in the face, and read the eye catching action.疋 Please refer to Figure 2 and Figure 3 together. Figure 3 is a flow chart of the image capture of the image taken. It comprises the following steps: 掏 Step 300: The first-photographing device 52 takes a picture of the face of the user. Step 310: Determining whether the eye portion of the facial image is located at a preset position of the facial image. In this embodiment, the determining ^56 determines that the eye portion of the facial image is the facial image. Central location. If yes, execute step_, if not, execute the bucket step 3〇4 ′ first-photographing device to capture the user's eye image. Step 306: The judging unit 56 judges whether the eye image is located at the center of the screen, "Step 8 and if not, execute step 312. Right step: Ship image "Image pre-processing unit 2". Step: Adjustment The movement adjustment is performed, and the shooting position 58 of the device 52 uses the stepping motor, so the first photography can be fine-tuned. The 52-position position configuration step 312 is adjusted to the fine structure 58: the position is set, the 58 series is used, and the stepping motor is used. Therefore, the photographing position of the second photographing device 52 is 200905577. After the photographing device 52 photographs the facial image (step 3〇〇), the judging unit & determines whether the eye portion is located at the center of the facial image. Since the pupil of the eye has a gray scale value which is extremely low, exhibiting a circular area and being located at the center of the person, the judging unit 52 will use the presence or absence of the pupil as the outline of the eye portion. First, the machine unit Μ compares with the 灰-photographing device 52 to capture all the pixel images of the face image—the critical gray scale value, when the gray level value of the image is larger than the critical reading, the minus pixel ash is set to 255. When the grayscale value of the pixel is less than the critical grayscale value, the grayscale value of the pixel is set to be such that the facial image presents a pixel binarized image.凊-Parallel M 4A_4C ® 1 4A_4D Schematic diagram of the radius of the broken pupil. After the face image captured by the first camera device 52 is binarized, an image as shown in the second figure is presented. Since the pupil has an approximately circular shape, the following will be referred to as a rounding mechanism to determine the position of the pupil. Since the pupil is the region with the lowest pixel value in the human eye image, the boundary of the pupil is the place where the grayscale value changes the most. That is, calculate the difference between the sum of the pixel values of the points on the circumference of the inner circle (radius mi) and the sum of the pixel values of the points on the circumference of the outer circle (radius r〇=r+1), if the difference is the largest The circular area determines that the circular area is the position of the pupil. For example, the judging unit 56 establishes a circular mask with the inner and outer diameters being the same as I' and r〇, where 内ι<1〇, r〇>3〇. On the circumference of η, take 1 pixel point A1 Α2, ·.., Α10', and take 1 pixel point Β1, Β2, ..., Β1〇 on the circumference of r〇, then the pixel on the inner circumference point Value corpse /=jP(10)+ Ρ(10)+ · + ρ(·) and the sum of the pixel values on the outer circumference point is /W(four)+ _) + ··. + />_. And calculate the difference between the outer circle and the inner circle pixel Ρ = 卜 For the pupil within the size range, the highest difference will be obtained (such as the 4th (: Fig. 200905577 not). When the transfer element 5_ has the highest The lang, the job department contains part of the eye area, and the judgment unit determines whether the pixel point is at the center of the face image. If the f camera is shooting, the user's face is not evenly photographed. If the eye, or part of the eye, is not in the middle of the face image, the adjustment mechanism % will adjust the face image of the first camera unit until the judgment unit determines that the face image t ^ has eyes and eyes. It is located at the position between the face and the face of the towel. Once the judgment unit 56 __ contains the _ field, and _ 〇 is located at the _image _ _, the second _ set 54 will _ - take the 52 determined position to shoot the user's eye u Preferably, the number of pixels of the second photographing device 54 is more than that of the second photographing device 54 and the image quality is better, so that the second photographing device 54 strongly captures an eye image with better resolution. Judging the set of the above-mentioned rounding mechanism to determine the second photography 54 + • ^ Set 54 the eye of the eye image is located in the center position. If the pupil is located in the center of the eye image 彳 4 靡忒 eye image 5 is transmitted to the image before the 〇 8 for the next step - processing; ^ Otherwise, the adjustment mechanism 312 adjusts the shooting position of the second photographing device 54, and photographicly captures the user's eye image again until the judging unit 56 judges that the pupil is located at the middle of the eye image. i see Fig. 5 'Fig. 5 The image pre-processing module is shown in the function block diagram. The image pre-processing her 2Q includes - the pupil positioning material 22, the rainbow position unit ^, an iris image navigation unit 26 and an image enhancement unit. After the eye image $ is taken out, the image pre-processing module (4) the fine pupil positioning unit 22 positions the pupil of the eye image $ to determine the center of the eye image. The private position unit Μ according to the center of the eye image to the eye 12 200905577 The iris boundary of the eye image 5 is positioned to mark the annular region where the iris is located. Then, the iris image normalization unit 26 is used to normalize the inner position of the iris and the object, and the iris image is hidden. Zoom to have the phase_inner and outer diameter size. In order to improve the recognition effect, the image information is enhanced to enhance the texture information in the rainbow trout. The image enhancement unit 28 only requires the insect film and the pupil portion due to the input eye image 5. Image, the position of the pupil is not fixed at the center of the eye image 5. The iris is surrounded by the circular shape of the pupil, so to judge the position of the iris, the pupil verification must be determined first. When the pupil position is earlier than π 22, all the image _ of the image 5 is compared with the critical gray scale value. When the gray scale value of the pixel is greater than the critical gray scale value, the gray scale value of the pixel is set. In other words, conversely, when the pixel-like gray_small_critical system__, the grayscale value of the setpoint point is set to the heart--the eye image will present a pixel binarized eye image. Because the radius of the lying hole is 4-30 pixels, and the pupil is the area with the lowest pixel value in the human eye image towel, the boundary of the pupil is the place where the grayscale value changes greatly. That is to say, calculate the sum of the pixel values of the points on the circumference of the radius 与 and the difference between the total values of the pixel values of the points on the outer circle (radius +1). If there is a large difference in the path f ((4)) /2 _ area (four) _ round area is the position of the pupil. For example, the pupil positioning unit 22 establishes a circle with the pixel point q as the center, and two inner and outer regions respectively, and a concentric circumferential mask, wherein, alum. In the circle of ^ 〇 pixels A1, A2 '..., A10, 1 pixel is also taken on the circumference of r〇
Bl〇，則内圓周上邊點像素值巧=p(A) + p(j2) +…+户(4〇)以及 外圓周上邊點像素值總和為p。= p㈣+取Q + ‘ + 。並計算外圓以及内 13 200905577 圓像素總合之差p = |p ―沪丨。璧+ 7 '大小範圍之内的瞳孔，會得到最高的差 值(如第4C圖所示）。當瞳孔定位 π 22判斷具有敢鬲的差值時，即認定以 像素點0為一估計的瞳孔圓心。 請參閱第6圖’第6圖絲計圓心示意圖，陰影部份為可能的圓心位 置。在顧二值化目_彡像求得瞳紐計圓心之後，瞳孔定位單元22會將 原先的眼睛影像5觸1G轉她域⑽定為實關心可能位置。接下 來將E域8〇内所有的像素作為候選的圓心，搭配半徑η冬㈣的圓 t罩在眼睛影像5中搜尋。接著計算㈣«像素差距最大的位置，即為 精準位置由於本實施例先運用二值化的眼睛影像找出瞳孔的估計 /置所在的區域8〇,接者·第二攝影裝置％拍攝的眼睛影像5找尋 貝際瞳孔圓雜置，所以可以大大減少搜尋範圍進而提升制的速度。 …當曈孔位置確定後，域定位單元24係用來侧虹膜的區域範圍。 因為每個人人眼憎孔與虹朗相對關係不盡相同，不一定都是同心圓， 所以偵測到的瞳關心位置，可能跟虹朗圓心位置有所差異。因此，求 得瞳孔圓心之後，虹財位單元24，並喃L定料元22職到的瞳孔圓 心二〇為圓Ο，料闕心、〇朋1G個像素半徑長肋的所有像素點作為 估4虹膜15心，亚以關半徑為％個像素，相半徑為個像素做為圓 遮罩，執行測圓機制以搜尋虹膜區域。 接下來，虹膜影像正單元％會依據求得的曈孔以及虹膜影像之邊 界，將虹膜影像自眼睛影像5中切割出來，得—環狀區域影像(由兩個不同 14 200905577 、之圓所構成)。-般來說，從不同人所操取之虹膜影像，具有不同的大小， 即使疋同-個人的虹膜影像，也可能會因為瞳孔的縮放以及拍攝距離的不 同而使内外徑的大小發生·。同時，瞳孔與虹膜大多時候都並非同心圓， 圓心的偏離程度也因人而異。因此虹膜影像正規化單“即是絲將不同 的嶋像5調整物的尺寸和對朗位置，從㈣除平移、縮放對於 虹膜辨識的影響。虹膜影像正規化單元26將曈孔圓心定為參考點，對於每 一點作座標賴轉換至極座標位置，展開成為_固定大小之矩形區塊。 瞳孔圓心％到虹膜邊界的距 請參閱第7圖，第7圖係虹膜正規化之示意圖。^為細方向，且 介於瞳孔邊界以及虹膜邊界之_點，即虹職域。令叫从邮分別 為瞳孔的圓心以及半徑，尽⑼為在0方向上， 離 則虹膜邊界上的一像素點巩〜，^)滿足： I = + i?L (0) X cos(^) 0) (2) (3) (4) +/?i(0)xsin(0) 因此可以得到fl = 〜= 2 )2 + (^5 - j；B )2 = |〇sfi| = 從式(1)、（2)、（3)可以推導出义⑼：Bl〇, the pixel value on the inner circumference is ==p(A) + p(j2) +...+ household (4〇) and the sum of the pixel values on the outer circumference is p. = p (four) + take Q + ‘ + . And calculate the difference between the outer circle and the inner 13 200905577 round pixel sum p = |p ― Humin. The pupils within the 璧+ 7' size range will get the highest difference (as shown in Figure 4C). When the pupil position π 22 judges that there is a discrepancy difference, it is determined that the pixel point 0 is an estimated pupil center. Please refer to Figure 6 for the schematic diagram of the center of the circle of Figure 6. The shaded part is the possible center position. After the binarization target is obtained, the pupil positioning unit 22 sets the original eye image 5 to 1G to the other field (10) as the real care possible position. Next, all the pixels in the E-domain 8〇 are used as the candidate center, and the circle image of the radius η winter (four) is searched for in the eye image 5. Then calculate (4) «The position with the largest pixel difference, that is, the precise position. Since this embodiment first uses the binarized eye image to find the area where the pupil is estimated/positioned, the eye of the second camera device is taken. Image 5 finds the miscellaneous miscellaneous hole, so it can greatly reduce the search range and increase the speed of the system. ... When the pupil position is determined, the domain positioning unit 24 is used for the area range of the side iris. Because each person's eye pupils and Honglang's relative relationship are not the same, not necessarily concentric circles, so the detected position of care may be different from the position of the Honglang center. Therefore, after the pupil center is obtained, the rainbow position unit 24, and the pupil center of the pupil unit 22 is the circle, and all the pixels of the 1G pixel radius long rib of the heart and the heart are estimated. 4 The iris 15 heart, the sub-radius is a radius of 100 pixels, the phase radius is a pixel as a circular mask, and a circular measuring mechanism is performed to search for the iris region. Next, the positive image unit of the iris image will cut the iris image from the eye image 5 according to the obtained pupil and the boundary of the iris image, and the image of the annular region (consisting of two different 14 200905577 circles) ). In general, iris images taken from different people have different sizes. Even with the same-personal iris image, the size of the inner and outer diameters may occur due to the scaling of the pupil and the difference in shooting distance. At the same time, the pupil and the iris are not concentric at most, and the degree of deviation of the center varies from person to person. Therefore, the iris image normalization sheet "is the effect of the size and the lang position of the different 嶋5 调整 , , , , , 平移 平移 平移 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹 虹Point, for each point of the coordinate conversion to the polar coordinate position, expand into a rectangular block of _ fixed size. Please refer to Figure 7 for the distance from the center of the pupil to the iris boundary. Figure 7 is a schematic diagram of the normalization of the iris. Direction, and between the pupil boundary and the iris boundary, that is, the rainbow occupation field. Let the call from the post be the center and radius of the pupil, and (9) be in the 0 direction, away from the pixel on the iris boundary ~, ^) Satisfaction: I = + i?L (0) X cos(^) 0) (2) (3) (4) +/?i(0)xsin(0) Therefore, we can get fl = ~= 2 ) 2 + (^5 - j;B )2 = |〇sfi| = From (1), (2), (3), we can derive the meaning (9):
Rl(0) = \〇pB\ = axcos0 + ^Rs2-a2 sinθ2 正規化時，在每僻方向上，利用物以m率，將以每一點 映射至展開矩形中之㈣位置似）。這種映射對於平移和内外圓環的大小 15 200905577 變換具有不變性。 正規化後的影像會傳送至影像強化單元28進行等化處理。因為虹膜區 域内紋理_比度小，也就是說，虹酿域各像权灰階值會分佈在某— 段比較小_。因此影像強化單元28會將虹廳__，其使得所 有灰階值出現的概率相同。如_所示，第_、影像強化單元28執行 等化的方法流程圖，其步驟如下：Rl(0) = \〇pB\ = axcos0 + ^Rs2-a2 sin θ2 In normalization, in each secluded direction, the object is mapped at the m rate, and each point is mapped to the (four) position in the expanded rectangle. This mapping is invariant to the translation and the size of the inner and outer rings 15 200905577. The normalized image is sent to the image enhancement unit 28 for equalization. Because the texture _ ratio in the iris region is small, that is to say, the gray scale values of the imagery in the rainbow region are distributed in a certain segment. Therefore, the image enhancement unit 28 will illuminate the rainbow hall __, which makes the probability of occurrence of all gray scale values the same. As shown in FIG. 3, the image enhancement unit 28 performs a method flow chart of equalization, and the steps are as follows:
步驟800 : 列出虹膜區域影像物 WHi。其中L是灰階級 數。 步驟802 :統計各灰階值的像素個數吔），卜从"“小 步驟隊糊始影像直方圖各灰階級的出現頻率勢心…㈣ 為虹膜區域影像的全部像素個數。 步驟806 :計算累計分布函數。Step 800: List the iris area image object WHi. Where L is the number of gray classes. Step 802: Count the number of pixels of each grayscale value ,), and from the "" small step team paste image histogram, the appearance frequency of each gray level... (4) is the total number of pixels of the iris region image. Step 806 : Calculate the cumulative distribution function.
步驟 808 :應用 gi =int[255 數符號。 (,)+0.5]计算映射後的輸入影像的灰階級 P為輸出影像灰階級的個數，其中加為取整 ’獲得等化後的輸出 步驟⑽：用㈣的投射關係調整原始影像的灰階級 虹膜影像。 :/月㈣參閱第1圖。特徵抽取處理模組30包含-小波轉換單元32，係 徵*/、膜衫像應用小波轉換法(wavelet)抽取虹膜影像内的主要特 徵]波轉換—般被應用在一維的訊號處理或是二維的影像壓縮上，作法 16 200905577 都疋個錢轉換後將等倾的虹麟像分解成高頻與低躺部分。高頻 i虎包s邊緣資訊或是雜訊，低頻訊制近似於等化後的虹膜影像，所以 特徵抽取處理她3〇會捨鱗高頻訊躺訂侧織❹速傳輸速度。 最後，特徵比對模組4〇採用的是支援向量機(Supp〇rt Vector嶽㈣ …’’支比對模組40 t將特徵抽取處理模，组3〇所抽取的虹膜影像内的 +要特徵〜板型胃鄉5G⑽所齡的減筆虹赚料作輯，若比對相 、、八使用者身刀符合模型資料庫50儲存的資料，如此一來，虹膜辨 識系”先1G即可用來依據不同使用者的虹膜辨識其身份。 相較於先前技術，本發明係利用二值化眼睛影像的方式預先估計瞳孔 位於使用麵㈣像祕雜置，接下來概縣的目畴雜配合瞳孔估 十置來决速搜*瞳孔的圓心實際位置。之後，由瞳孔圓心的實際位置以 測圓機制找出虹麵域的範圍。透過上述機制，本發明可以減少搜尋瞳孔 圓。位置的時間’增加虹贿齡統職正雜和資料處理效率。這將有 助於未來虹_勒_龍化使用。 」本1月已用|x佳實施例揭露如上，然其並翻以限定本發明，任 D /¾匕技*者，在不脫離本發明之精神和範圍N，當可作各種之更動與 修改’因此本發日月之保護範圍當視後附之申請專利範圍所界定者為準。 【圖式簡單說明】 第1圖係本㈣之虹__統之功能方塊圖。 17 200905577 第2圖係第1圖之影像擷取裝置之功能方塊圖。 第3圖係第1圖之影像擷取裝置擷取眼睛影像之流程圖。 第4A-4C圖係判斷瞳孔半徑範圍之示意圖。 第5圖係第1圖之影像前置處理模組之功能方塊圖。 第6圖係估計圓心示意圖 第7圖係虹膜正規化之示意圖。 第8圖係影像強化單元執行等化的方法流程圖。 【主要元件符號說明】 10 虹膜辨識系統 50 影像擷取裝置50 20 影像前置處理模組 30 特徵抽取處理模組 40 特徵比對模組 50 模型資料庫 52 第一攝影裝置 54 第二攝影裝置 56 判斷單元 58 調整機構 55 紅外線投射器 80 區域 32 小波轉換單元 18Step 808: Apply gi =int[255 number sign. (,) +0.5] Calculate the gray level P of the mapped input image as the number of output image gray levels, where the addition is rounded to obtain the equalized output step (10): adjust the gray of the original image with the projection relationship of (4) Class iris image. : / month (four) see Figure 1. The feature extraction processing module 30 includes a -wavelet conversion unit 32, the signature */, the film image is applied by wavelet transform to extract the main features in the iris image] the wave conversion is generally applied to the one-dimensional signal processing or In the two-dimensional image compression, the practice 16 200905577 all converts the isoping Honglin image into high frequency and low lying parts after a money conversion. High-frequency i tiger bag s edge information or noise, low-frequency communication system is similar to the iris image after equalization, so the feature extraction process will be 3 〇 〇 高频 高频 高频 高频 高频 高频 高频 高频 高频 高频 高频 高频 高频 高频 高频 高频 高频 高频 高频 高频 高频 高频Finally, the feature comparison module 4〇 uses a support vector machine (Supp〇rt Vector Yue (4) ... ''sense comparison module 40 t to extract the feature processing mode, the group 3 〇 extracted iris image + Feature ~ plate type stomach town 5G (10) age of the pen to reduce the amount of energy, if the phase, eight user body knife meets the data stored in the model database 50, so that the iris identification system can be used first 1G In accordance with the prior art, the present invention uses the method of binarizing the eye image to pre-estimate that the pupil is located on the use surface (4), and then the pupil of the county is mixed with the pupil. It is estimated that the actual position of the center of the pupil is searched. After that, the actual position of the center of the pupil is used to find the range of the rainbow surface by the circle measuring mechanism. Through the above mechanism, the present invention can reduce the search for the pupil circle. Increasing the efficiency of data processing and data processing. This will help the future use of rainbow _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Anyone who is D / 3⁄4 匕 * The spirit and scope of the present invention, N, can be varied and modified. Therefore, the scope of protection of the present invention is subject to the definition of the patent application scope. (4) The function block diagram of the rainbow __ system. 17 200905577 Fig. 2 is a functional block diagram of the image capture device of Fig. 1. Fig. 3 is a flow chart of the image capture device of Fig. 1 for capturing the eye image. Fig. 4A-4C is a schematic diagram for judging the radius of the pupil. Fig. 5 is a functional block diagram of the image preprocessing module of Fig. 1. Fig. 6 is a schematic diagram showing the normalization of the iris by the schematic diagram of the center of the figure. 8 is a flow chart of the method for performing equalization of the image enhancement unit. [Main component symbol description] 10 Iris recognition system 50 Image capture device 50 20 Image pre-processing module 30 Feature extraction processing module 40 Feature comparison module 50 Model database 52 first photographing device 54 second photographing device 56 judging unit 58 adjusting mechanism 55 infrared projector 80 region 32 wavelet transform unit 18