1298852 九、發明說明： 【發明所屬之技術領域】 本發明係關於一種連續性指紋影像擷取裝置及其 處理方法，尤其關於一種利用滑動式指紋感測器擷取 連續性片段指紋影像以及特徵點幾何座標疊合的方式 來元成指紋登入及後續辨識的處理方法。本發明係部 份關聯到本案其中一個發明人之下述專利申請案：i 中華民國發明專利申請案號〇91106806，申請日為2〇〇2 年4月3日，發明名稱為「電容式指紋讀取晶片」， 公告號為583592:2.中華民國發明專利申請案號 〇9〇112〇23,申請曰為民國2001年5月17曰，發^ 名稱為「電容式壓力微感測元件及其製造方法與訊號 :賣取方式」，現已獲得專利，其證書號碼為發明專利 第182652 ;及3.中華民國發明專利申請案號 〇901 13755,申請曰2001年7月6曰，發明名稱為「溫 差感應式指紋辨識晶片之設計與製造」，公告號為 558686，證書號碼為發明第189671號。 … 【先前技術】 習知的指紋讀取方；去，可以利用沾冑墨水之手浐 按遂在紙上，再利用光學掃描器來掃描該紙張以將指曰 紋輸入電腦中’然後與資料庫中之指紋圖形比對。然 而’上述方法之最大缺點為無法達到即時處理的目 的，因此無法滿足越來越多即時認證的需求如： 網路認證，電子商務，攜帶式電子產品保密，κ卡個 Ϊ298852 人身分認證，保全系統等等。 即日守的指紋讀取方法便成為生物辨識市場中的關 鍵技術。傳統上，可使用光學式指紋感測器來即時^ 取指紋，然而其具有體積過於龐大之缺點。為此，= 用矽半導體的晶片式指紋感測器因應而生，克服了上 述光學式感測器的缺點，例如祥群科技之產品型號 LTT-C5 00電容式指紋感測器即有此優點。 由於手指的尺寸限制，習知的晶片式指紋感測哭 之感測面積至少大於9mm*9mm。再者，由於矽積二 電路製造之限制，在一片6吋晶圓中僅能產出5〇至 個有效晶片，再加上封裝測試成本，將使得單一指紋 感測器之售價至少大於美金10丨以上，這將心其 應用於各種消費性電子產品，譬如筆記型電腦、手機、、 個人數位助理，電腦周邊產m於整合指紋感測 器的個人身分識別卡。 為了克服上述成本問題，可以將傳統二維晶片式 指紋感測器之長度予以縮小，藉以增加有效晶片的產 出數目並降低晶片單價。其原理係藉由手指滑動過晶 片表面，完成整個手指的掃描，再將每張擷取到的片曰 段指紋影像建立關連性儲存成—參考模塊，後續的使 用也疋相同的將手指滑過晶片表®，並得到-指紋模 塊，透過指紋模塊與參考模塊的比較便可以確認使用 iVidmgUet在美國專利篪公報 及Kramer在美國直各丨（”， 秀闼專利6,3 17,508號公報分別揭露了 工298852 2之一種滑動式指紋感測器及多重影像串接的方法， 测二〇與2所示。_ 1所示為習知應用滑動式指紋感 D 作為手指指紋圖像讀取的架構示意圖。豆中 二：器10為-陣列元件，其橫轴方向尺寸約等；手 方？广而縱轴方向(手指滑動方向)尺寸遠小於 声方向士寸’手指20與感測器10有一相對運動速 =，也就是說將手指以速度V滑動過感測器1〇的 =面’因此感測器10可以連續性的掏取片段指紋圖 纹路°Ρ™2左邊所示即㈣擷取到的連續性片段手指 、圖像21)。這些連續性片段手指紋路圖像21可以 m至-微處理器3G’並且暫存於_隨機存取記憶 (AM)40中。隨後’微處理器3〇將這些連續性片 L手μ指紋路圖像21取出，藉由儲存於唯讀記憶體 (〇M)5G巾的軟體邏輯（algGrism)做手指紋路影 合動作，亦即將複數張的片段手指紋路圖像21接合 成-張完整手指紋路圖像22’也就是如圖2右邊所： 的指紋全圖。1298852 IX. Description of the Invention: [Technical Field] The present invention relates to a continuous fingerprint image capturing device and a processing method thereof, and more particularly to a method for capturing continuous segment fingerprint images and feature points by using a sliding fingerprint sensor The way in which the geometric coordinates are superimposed is used to process the fingerprint login and subsequent identification. The invention is partially related to the following patent application of one of the inventors of the present invention: i, the Republic of China invention patent application number 〇91106806, the application date is April 2, 2, 2, the invention name is "capacitive fingerprint Read the wafer", the announcement number is 583592: 2. The Republic of China invention patent application number 〇9〇112〇23, the application for the Republic of China on May 17, 2001, issued the name "capacitive pressure micro-sensing components and Its manufacturing method and signal: selling method" has been patented, its certificate number is invention patent No. 182652; and 3. Republic of China invention patent application number 〇 901 13755, application 7 July 6, 2001, invention name For the design and manufacture of temperature-sensitive inductive fingerprint identification wafers, the bulletin number is 558686 and the certificate number is invention No. 189671. ... [Prior Art] The conventional fingerprint reader; go, you can use the hand of the ink to press the paper on the paper, and then use the optical scanner to scan the paper to input the fingerprint into the computer. Then the database The fingerprint pattern in the comparison. However, the biggest disadvantage of the above method is that it can not achieve the purpose of real-time processing, so it can not meet the needs of more and more instant authentication such as: network authentication, e-commerce, portable electronic product confidentiality, κ card Ϊ 298852 identity authentication, security System and so on. The fingerprint reading method of the day is becoming a key technology in the biometric market. Traditionally, optical fingerprint sensors can be used to capture fingerprints in real time, however, they have the disadvantage of being too bulky. For this reason, the chip-type fingerprint sensor for semiconductors has been used to overcome the shortcomings of the above-mentioned optical sensors. For example, the product model LTT-C5 00 capacitive fingerprint sensor of Xiangqun Technology has this advantage. . Due to the size limitation of the finger, the sensing area of the conventional wafer type fingerprint sensing crying is at least greater than 9 mm * 9 mm. Furthermore, due to the limitations of the hoarding two circuit manufacturing, only 5 〇 to a valid chip can be produced in a 6-inch wafer, and the package test cost will make the price of a single fingerprint sensor at least greater than the US dollar. More than 10 inches, this will be applied to a variety of consumer electronic products, such as notebook computers, mobile phones, personal digital assistants, personal identification cards integrated with fingerprint sensors. In order to overcome the above cost problem, the length of the conventional two-dimensional wafer type fingerprint sensor can be reduced to increase the number of effective wafers produced and reduce the unit price of the wafer. The principle is that the finger is swept across the surface of the wafer to complete the scanning of the entire finger, and then the fingerprint image of each captured segment is stored and stored as a reference module, and the subsequent use is also the same. Wafer Table®, and get the - fingerprint module, through the comparison of the fingerprint module and the reference module, it can be confirmed that the use of iVidmgUet in the US Patent Gazette and Kramer in the United States (", Xiuyan Patent 6,3 17,508 respectively revealed A sliding fingerprint sensor of 298852 2 and a method for serially connecting multiple images, measured by two and two. _ 1 shows a schematic diagram of a conventional application of sliding fingerprint sense D as a fingerprint reading of a finger fingerprint. Bean 2: The device 10 is an array element, and its horizontal axis dimension is about the same; the hand is wide and the vertical axis direction (finger sliding direction) is much smaller than the sound direction. The finger 20 has a relative to the sensor 10 The speed of motion = that is, the finger is slid over the sensor's 1 〇 = face at the speed V. Therefore, the sensor 10 can continuously capture the segment fingerprint pattern. The left side of the TM2 is displayed (4). Continuous Fragment finger, image 21). These consecutive segment hand fingerprint road images 21 can be m to - microprocessor 3G' and temporarily stored in _ random access memory (AM) 40. Then 'microprocessor 3 〇 these The continuous film L-hand fingerprint road image 21 is taken out, and the fingerprint logic road moving action is performed by the software logic (algGrism) stored in the read-only memory (〇M) 5G towel, that is, the fragment hand fingerprint road image is also a plurality of images. 21 is joined into a complete hand fingerprint road image 22', that is, the full fingerprint of the fingerprint on the right side of Fig. 2.
這種方法的優點是可以擷取到相當大的手指紋路 圖像，卻不需要大面積的感測器，不僅在成本上有相 當大的優勢’辨識比對的品質也如大面積指紋感測器 般的好，例如低誤判率及聚判率（false rejecti⑽U 然而，這種手指紋路圖像的處理方法也有一2缺 點。首先，必須要擷取非常多的片段手指紋路圖像2一i，、 其通常需要百張甚至更多的片段手指紋路圖像。舉例 而言’若指紋感測器的規格為8*28〇(此為 7 1298852The advantage of this method is that it can capture a large hand fingerprint image, but does not require a large area of the sensor, not only has a considerable cost advantage. 'The quality of the identification is also like the large area fingerprint sensing. Good like, such as low false positive rate and multi-judgment rate (false rejecti (10) U However, this method of processing hand-finished road images also has one disadvantage. First, it is necessary to capture a lot of segment fingerprint road image 2 i, It usually requires a hundred or more segments of the fingerprint road image. For example, 'If the size of the fingerprint sensor is 8*28〇 (this is 7 1298852)
Fingerchip的規格）’則隨機存取記憶體4〇至少需要 約3〇〇kbyte的容量（再加上後續的接圖處理的緩^記 憶體（buffer memory)可能需要更多），因而增加了系絲 設計時的成本。 ’' 再者，每個人手指滑動時在不同狀況下所施的壓 力不同，因此所接合成的手指紋路圖像通常在縱車由方 向會有不一的變形，不利於後續的辨識比對，特別是 應用於不同指紋感測器間的系統比對，例如網路的： 端分別使用滑動式及面積型指㈣測器，％有可 法相互比對。 匕…、 ，〜π π々仅于指 、、文路圖像都需要利用影像處理做全圖比對，以找出曰 小變異量作為接圖的基準，此種方法例如習知影像= =處理中的區塊比對（blGek她hing)，此種方法是相 田費日守的’其通常需要功能強大的微處理器30來處 理丄在系統設計時亦將增加成本，抑或增加處理時： 而谷易造成使用者的不冑。而且手指滑動時的方向偏 移更是需要做複雜的旋轉接圖考量，且通常容易失 敗0 其他的解決方法，例如Svens_等人在美國專利 細伽中揭露的方法捨棄了上述的指紋 圖像接合的方法。簡述之，在第一次指紋登入過程中， 存取了複數張片段指紋圖像的特徵區&，以建立一組 f考比較資料表’在辨識比對時所擷取的任何一張片且 長指紋圖像都立即地與該組參考比較資料表作交又比 1298852 K然其號稱可以減少上述隨機存取記憶體40的 求’ “卻增加唯讀記憶體提供參考指紋模板 為°亥、且參考比較資料表）唯讀記憶體容量的需 求’ 一樣需要增加成本。 因為母_人在比對時所操取到的任一片段指 :資料與該組參考比較資料表之間都是獨立的行為， =不像接圖叙的具有—對—關係，亦即是隨機的行 2，也就是說可能登入時所筛選的複數張片段指紋特 :政區塊與後續比對的片段指紋圖像並無任何重疊關 連，因此常需要使用者重複滑動手指方能通過認證， 此舉增加使用者不便及降低使用意願。 並且，在比對時每―片段指紋圖像都需要與該組 參考比較貢料表(含複數張片段指紋資料)作交叉比 對，因而需要強大的資料運算能力的微處理器30方 能解決，亦不利於成本之降低。 【發明内容】 因此，本發明之-個目的係一種利用 感測器操取連續性片段指紋影像以及特徵點幾何二 疊合的方式來完成指紋登入及後續辨識的處理方法： 可:統設計時的隨機存取記憶體、唯讀記憶體 的今里需纟，且降低微處理器效能之需求 降低最終指紋擷取裝置之成本。 精乂百效 本發明之另一個目的将挺祝 _ ^ 刃係k供一種將連續性 紋圖像即時萃取出片段特徵點矩陣以建立特徵點幾； 1298852 21之方法’此舉不需要複雜的浮點運算功能 由硬體的輸出邏輯完成。 u 饩本發明之又一目的係提供-種建立連續性片段特 :點幾何座標資料關連性的方法，以疊合成一完整： 特徵點矩陣資料’提供手指登人及後續比對使用。 本發明之再一目的係提供一種建立連續性片 相幾何座標資料關連性的方法， 特徵點矩陣資料’幾何特徵點座標的疊丄 滑動時方向的改變及旋轉所造成之影響。 曰、 為達上述目的，本發明提供—種連續性指紋影像 。裝置，其包含一滑動式指紋感測器、-輸出邏輯、 —微處理器、一唯讀印悟獅s ,左a 一 u 羋貝口己隐體及一隨機存取記憶體。滑 式指紋感測器用以感测滑動通過其上之一手指之、— ί個片段指紋圖像。輸出邏輯連接至滑動式指：感： -’用以接收該等片段指紋圖像’並將該等 圖像即時萃取出複數個片段特徵點矩陣，且依㈣等 片段特徵點矩陣建立片段特徵點幾何座標資料。微處 王。益連接至輸出邏輯。唯讀記憶體中儲存有一軟體邏 輯。:亚連接至微處理器。隨機存取記憶體連接至微處 理益’用以暫存片段特徵點幾何座標資料。微處理哭 :據軟體邏輯將該“段指紋圖像之該等片段特㈣ 何座標貧料建立關連性以疊合成含有-完整特徵點 矩陣之完整特徵點幾何座標資料。 上述之輸出邏輯亦可由儲存於唯讀記憶體中之另 -軟體邏輯來取代，且本發明亦提供一種連續性指紋 10 1298852 影像擷取裝置之處理方法，包含以下步驟··擷取滑 通過-滑動式指紋感測器上之一手指之複數個片： 紋圖像’·將該等片段I紋圖像即時萃取出複數個片又严曰 特徵點矩陣’·依據該等片段特徵點矩陣建立片段特： 點幾何座標資料並暫存之；及將該等片段指紋圖像： 該等片段特徵點幾何座標資料建立關連性以疊合成含 有70整特彳政點矩陣之完整特徵點幾何座標資料。— 、猎由上述之裝置與方法，可以有效降低擷取裝置 之成本，並解決習知技術之影像疊合的複雜問題。 【實施方式】 圖3顯示依據本發明第一實施例之連續性指紋影 ^擷取裝置的架構示意圖。圖4所示即為利用圖3之 架構所擷取到的連續性片段指紋圖像、即時片段特徵 ”、、占成何座標貧料萃取以及疊合成之完整特徵點幾何座 標資料之示意圖。如圖3與4所示，本實施例之連續 2指紋影像擷取裝置包含一滑動式指紋感測器1〇〇、、 一輸出邏輯600、一微處理器3〇〇、一唯讀記憶體 (R〇M)5〇〇及一隨機存取記憶體（RAM)4〇〇。 滑動式指紋感測器100為一陣列元件，其基本上 包含一基板、一感測元陣列及一周邊處理及控制電 路、感測7L陣列係位於基板上，並由複數個指紋感測 :=二維陣列的方式排列所構成。指紋感測元可以是 電容式、壓力式或溫差式感測元，如本案申請人於下 述專利申請案所揭露的，在此不贅述·· 11 1298852 中華民國發明專利申請案號091 106806，申請 日為2002年4 R 1 々 月 取曰p 月3日，發明名稱為「電容式指紋讀 取曰曰片」，公告號為583592; 中華民國發明專利申請案號0901 12023，申嗜 曰為民國2001牟s曰17 α 々 年月17日，發明名稱為「電容式壓 刀破感測元件及Α _、生 千及，、製仏方法與訊號讀取方式」，現已 又传利’其證書號碼為發明專利第丨82652號，·及 中華民國發明專利申請案號〇9〇1 13755，申請 2 2〇01年7月6曰，發明名稱為「溫差感應式指紋 辨硪晶片之設計與製造」，公告號為558686，證書號 碼為發明第189671號。 玲儿 、其中在本實施例中，感測器所採用的原理係電容 式’感 >則7L陣列規格為橫向彳2〇〇個_元晝素，縱 向則有至少16個畫素以上，否則不易在單一片段指 紋圖像中萃取出特徵點資料，而每一感測元包含了2 感測電極及位於其下方基板中的一感測電路，每—晝 素的尺寸為解析度為5〇8dpi。 滑動式指紋感測器100係用以感測滑動通過其上 之一手指200之複數個片段指紋圖像2丨〇，譬如是片 段灰階指紋圖像。輸出邏輯6〇〇可以整合於滑動2指 紋感測器100中，或可以與滑動式指紋感測器^㈧呈 分離狀。輸出邏輯600係連接至滑動式指紋感測= 100，用以接收該等片段指紋圖像210，並將該等片严 指紋圖像210即時萃取出複數個片段特徵點矩= 220 ’且依據該等片段特徵點矩陣22〇建立片段特1 12 1298852 點幾何座標詞230。微處理器係連接至輪 輯600。唯讀記憶體500中儲存有一軟體邏輯51〇, 且唯讀記憶體500係連接至微處理器3〇〇。隨機存取 記憶體400係連接至微處理器3〇〇 ’用以暫存片段 徵點幾何座標資料230。 ^ 在處理過程中，微處理器300依據軟體邏輯5ι〇 將該等片段指崎210之該等片段特徵點幾何座標 育料23G建立關連性，在本發明中係以疊合成含有一 完整特徵點矩$ 24G之完整特徵點幾何座標 250。 ^ 、2實施例有別於習知手指紋路接合技術的特色 為田手私2 〇 〇滑動過感測器表面時，所擷取到的片 &灰k扎紋圖像係透過輸出邏輯6⑽即時的將該灰階 :像直接萃取出特徵點矩陣，如圖4之標號所示， 每張輸出的圖像在暫存於隨機存取記憶體400之前 都已、、二被轉換成片段特徵點幾何座標資料230 (圓圈 、斤示之集口），其特色為每一片段特徵點幾何座標資 料230的資料量相當小（僅幾十個位元組沙 S此即使手指200滑過感測器表面擷取到100張片段 才曰、、文’轉換成片段特徵點幾何座標資料230輸出後暫 存於隨機存取記憶體400也只需要幾千個位元組 (kbytes)的空間即可，纟大的減少暫存記憶體的容量 需求，降低系統設計成本。 圖5顯不依據本發明第二實施例之連續性指紋影 像擷取裝置的架構示意圖。如圖5與4所#，本實施 13 1298852 例之連續性指紋影像擷取裝置包含一滑動式指紋感測 态100、一 Μ處理态300、一唯讀記憶體5〇〇、一隨機 存取記憶體400。上述之構件與作動方式係類似於第 一實施例，以下僅說明與第一實施例不同之處。 U處理裔3 00接收複數個片段指紋圖像2丨〇，唯 唄纪憶體500中儲存有一軟體邏輯51〇⑷。微 處理器300依據軟體邏輯51〇將該等片段指紋圖像21〇 即時萃取出複數個片段特徵點矩陣22〇，並依據該等 片奴特欲點矩陣220建立片段特徵點幾何座標資料 230並暫存於该隨機存取記憶體4⑽中，然後將該等 片奴才曰紋圖像2 1 0之該等片段特徵點幾何座標資料 230建立關連性以疊合成含有一完整特徵點矩陣24〇 之70整特徵點幾何座標資料2 5 0。 ^在上述兩實施例中，儲存於唯讀記憶體500中的 ^體邏輯510或輸出邏輯600可提供微處理器3〇〇對 母一片段指紋圖像作即時特徵萃取及建立幾何座標資 ，的運算邏輯。由於本發明實施例係將灰階圖像直接 萃取出特徵座標，％全不需要如傳統般的處理程序， ，如灰階指紋圖像、影像前處理及二階化、細線化、 最後：徵萃取。因此不需要影像前處理及二階化的複 雜，开步驟，因此可以大大的節省該微處理器300的 ，才^力需求。也因為這樣的特徵萃取方式相當簡 : 可以直接以一硬體化的輸出邏輯600執行此一功 月匕口’该輪出邏輯本身也不需要具有浮點運算功能，因 此可以輕易的整合於指紋感測器之單晶片中。 14 1298852 如圖3至5所示，每一片段指紋圖像都被即時處 理成片奴特徵點幾何座標資料23〇，而全部暫存於隨 機存取記憶體400巾。唯讀記憶體5〇〇中的軟體邏輯 (aborism)亦提供一片段特徵點之幾何座標疊合運算 邏輯，將所有的片段特徵點幾何座標資料230疊合成 完整特徵點幾何座標資料250，如圖4之右邊所示。 圖6顯示本發明之連續性指紋影像擷取裝置之處 理方法之後私圖。如圖6所示，本發明之處理方法包 含以下步驟： /驟S 1 〇 ·開始，其可以是開始指紋登入或開始 比對扣紋，於開始登入指紋的情況下，唯讀記憶體5⑻ 並’又有儲存有參考模板，或者是唯讀記憶體5原本 所儲存的參考模板將被清除，而於開始比對指紋的情 況下，唯讀記憶體500中已經儲存有參考模板； 步驟S20··擷取滑動通過一滑動式指紋感測器ι〇〇 上之一手指200之複數個片段指紋圖像21〇 ; 卜步驟S30 :將該等片段指紋圖像210即時萃取出 複數個片段特徵點矩陣220 ; 步驟S40 :依據該等片段特徵點矩陣22〇建立片 段特徵點幾何座標資料23〇並暫存之； 步驟S50:將該等片段指紋圖像21〇之該等片段 特徵點幾何座標資# 23〇建立關連性以疊合成含有一 完整特徵點矩陣240之完整特徵點幾何座標資料 250。其中，於指紋登入模式下，可以將完整特徵點 幾何座標資料250所形成的參相板料於唯讀記憶 15 1298852 體中；或者於比對楹4 ΠΓ 1 挪一 、工下，可以將完整特徵點幾何座 標貧料250與儲存於唯靖 % "貝"己丨思體中的苓考模板比對， 以確認此使用者是否為授權使用者；及 步驟S 6 0 ··結走，在々本土上 .^ 係、、Ό束扎紋登入或指紋比對的 勳作。 將片段特徵點幾何座標資料疊合的優點，除了上 =可以節省隨機記憶體的需求空間以利降低系統設計 、本卜纟可以降低運算的複雜度，因為習知技術中 的手指紋路圖像接圖，需要相鄰兩張片段手指紋路圖 像做全圖比對找出最小變異量作為接圖的基準，習知 方法都是影像處理常採用的區塊比對（bi〇ck 她㈣），此種方法運算量大且費時。而本發明實施 例所採用的方法，才目鄰兩張幾何座標資料僅有少數的 特徵點幾何座標需要作疊合的邏輯運算，因此比較於 全圖接合，運算量小且省時。 且幾何座標的疊合對於旋轉也是相當容易的， ’、疋兒如果手扣滑動時有方向轉動的情況發生 時/習知技術的手指紋路接圖方式是很難解決的，因 為第二張圖如果與第一張圖存在一角度時，（例如感 測器規格為8*280此為Atmel Fingerchip規格），則 ”要者之間角度差異1度以上，兩張片段指紋圖像 便有部份區域無法重疊，造成接圖的困難發生，這是 習知技術很難突破的技術瓶頸。然而對於本發明實施 例而5，特徵點幾何座標的旋轉仍可以輕易完成疊合 的動作，而且不會遺失手指資訊。 16 1298852 与傻Γ人手指滑動時的速度及壓力不同所導致的指紋 二、:形問題’習知採用手指紋路接圖的方法，每— 、兄' 及：Ϊ紋圖像所受不一的壓力係來自不同的使用情 路=動時的非定速行為，因此所接合出來的手指紋 圖像4伸長或縮短的情況發生’亦或者在一部份 二：騎*，-部份區域為縮短的情況發生，而不利 績的辨識比對或不同感測器系統間的相互認證，Fingerchip's specification) 'The random access memory 4〇 needs at least 3〇〇kbyte capacity (plus the buffer memory of subsequent connection processing may need more), thus increasing the system. The cost of silk design. '' Furthermore, each person's finger is different in pressure under different conditions, so the image of the hand-finished road that is joined is usually deformed in the direction of the longitudinal direction, which is not conducive to subsequent identification comparison. In particular, it is applied to system comparison between different fingerprint sensors. For example, the end of the network uses sliding type and area type (four) detectors respectively, and % can be compared with each other.匕..., ,~π π々 Only for the image of the finger and the text, it is necessary to use the image processing to make a full map comparison to find the small variation as the reference for the connection. This method is for example, the conventional image == In the process of block comparison (blGek she hing), this method is the same as the "study of the field", which usually requires a powerful microprocessor 30 to handle the increase in cost when the system is designed, or increase processing: And Gu Yi caused the user's innocence. Moreover, the direction shift when the finger slides is more complicated rotation diagram consideration, and it is usually easy to fail. 0 Other solutions, such as the method disclosed by Svens_ et al. in U.S. Patent Saga, discard the above fingerprint image. The method of joining. Briefly, during the first fingerprint login process, the feature area & of a plurality of segment fingerprint images is accessed to establish a set of f-test comparison data table 'any one captured during the identification comparison The long and long fingerprint images are immediately intersected with the reference data table of the group and compared with the 1289852 K. It is claimed that the random access memory 40 can be reduced. Hai, and refer to the comparison data table) the requirement of read-only memory capacity' needs to increase the cost. Because the parent_person is in the comparison when any segment is taken: the data is compared with the reference data table of the group. It is an independent behavior, = unlike the connected-to-relationship relationship, that is, the random line 2, that is, the multiple fingerprints that may be screened when logging in: the political block and the subsequent comparison The segment fingerprint image does not have any overlapping relationship, so it is often required that the user repeatedly slides the finger to pass the authentication, which increases the user's inconvenience and reduces the willingness to use. Moreover, each of the segment fingerprint images needs to be compared with the image. Group The comparison of the tribute table (including the plurality of pieces of fingerprint data) for cross-comparison, so that the microprocessor 30, which requires a powerful data computing capability, can be solved, which is also disadvantageous for cost reduction. [Invention] Therefore, the present invention - A purpose is to use a sensor to capture a continuous segment fingerprint image and feature point geometry two superimposed way to complete the fingerprint login and subsequent identification processing methods: Can: design random access memory, read only The memory needs to be reduced today, and the need to reduce the performance of the microprocessor reduces the cost of the final fingerprint capture device. The other purpose of the invention is that it would be a good idea to provide a continuous pattern image. Instant extraction of the segment feature point matrix to establish feature points; 1298852 21 method 'This does not require complex floating-point arithmetic functions to be completed by the hardware output logic. u 又一 Another object of the present invention is to provide a continuous Sexual fragment: a method of correlation of geometric coordinates of a point, to synthesize a complete: feature point matrix data 'providing finger landing and subsequent comparison use. The invention A further object is to provide a method for establishing the correlation of continuous phase-phase geometric coordinate data, the characteristic point matrix data 'the change of the direction of the geometric feature point coordinate when sliding and the rotation are caused. 曰, in order to achieve the above purpose, The invention provides a continuous fingerprint image device, which comprises a sliding fingerprint sensor, an output logic, a microprocessor, a read-only imprinted lion, a left a-u, a beiguchi, and a hidden body. a random access memory. The sliding fingerprint sensor is configured to sense a fingerprint image of a finger that slides over one of the fingers. The output logic is connected to the sliding finger: sense: - 'to receive the same The segment fingerprint image 'and extracts the images from the plurality of segment feature point matrices in real time, and establishes the segment feature point geometric coordinate data according to the (four) segment feature point matrix. The micro-portion is connected to the output logic. A software logic is stored in the read-only memory. : Sub-connected to the microprocessor. The random access memory is connected to the micro-processing entity to temporarily store the segment feature point geometric coordinate data. Micro-processing crying: According to the software logic, the segments of the segment fingerprint image are connected to each other to establish a correlation property to superimpose the complete feature point geometric coordinate data of the matrix containing the complete feature point. The above output logic can also be The software-storage logic stored in the read-only memory is replaced, and the present invention also provides a method for processing the image capture device of the continuous fingerprint 10 1298852, including the following steps: · Sliding through-sliding fingerprint sensor A plurality of slices of one finger: a grain image '· Instantly extracts a plurality of slices of the segment I image and severely characterize the matrix of the feature points. · Create a segment based on the feature point matrix of the segments: Point geometric coordinates Data and temporary storage; and the fingerprint images of the segments: the geometric coordinates of the feature points of the segments establish correlations to synthesize the complete feature point geometric coordinates of the matrix containing 70 full-dimensional punctuation points. The device and the method can effectively reduce the cost of the capturing device and solve the complicated problem of image superposition of the prior art. [Embodiment] FIG. 3 shows the first aspect of the present invention. A schematic diagram of the architecture of the continuous fingerprint image capture device of an embodiment. Figure 4 shows the continuous segment fingerprint image captured by the architecture of Fig. 3, the feature of the instant segment, and the poor standard of the image. Schematic diagram of the complete feature point geometric coordinates of the material extraction and stacking. As shown in FIG. 3 and FIG. 4, the continuous fingerprint image capturing device of the embodiment includes a sliding fingerprint sensor 1 , an output logic 600 , a microprocessor 3 , and a read-only memory. (R〇M) 5〇〇 and a random access memory (RAM) 4〇〇. The sliding fingerprint sensor 100 is an array component, which basically comprises a substrate, a sensing element array and a peripheral processing and control circuit, and the sensing 7L array is located on the substrate, and is sensed by a plurality of fingerprints: The two-dimensional array is arranged in a manner. The fingerprint sensing element may be a capacitive, pressure or temperature difference sensing element, as disclosed in the following patent application by the applicant of the present application, and is not described here. 11 1298852 Republic of China Invention Patent Application No. 091 106806, The application date is 2002 4 R 1 々月曰 曰p 3, the invention name is “capacitive fingerprint reading 」 film”, the announcement number is 583592; the Republic of China invention patent application number 0901 12023, Shen 曰The Republic of China 2001牟s曰17 α On the 17th of the year, the invention was titled “Capacitive Pressure Knife Breaking Sensing Components and Α _, Shengqian, and the Method of Reading and Signal Reading”. The certificate number is invention patent No. 82652, and the Republic of China invention patent application number 〇9〇1 13755, application July 2, 2001, 6 invented, the invention name is "temperature difference sensing fingerprint identification chip design And manufacturing, the announcement number is 558686, and the certificate number is invention No. 189671. Linger, in which, in the present embodiment, the principle adopted by the sensor is a capacitive 'feel>, the 7L array specification is horizontal 彳 2 _ _ yuan ,, and the vertical direction has at least 16 pixels. Otherwise, it is difficult to extract feature point data in a single segment fingerprint image, and each sensor element includes 2 sensing electrodes and a sensing circuit located in the substrate below, and the size of each halogen element is 5 〇 8dpi. The sliding fingerprint sensor 100 is configured to sense a plurality of segment fingerprint images 2 滑动 sliding through one of the upper fingers 200, such as a segment grayscale fingerprint image. The output logic 6〇〇 can be integrated into the sliding 2 fingerprint sensor 100 or can be separated from the sliding fingerprint sensor ^ (eight). The output logic 600 is connected to the sliding fingerprint sensing=100 for receiving the segment fingerprint images 210, and extracting the slice fingerprint images 210 into a plurality of segment feature point moments=220′ and according to the The segment feature point matrix 22 〇 creates a segment feature 1 12 1298852 point geometry coordinate word 230. The microprocessor is connected to the wheel 600. The read-only memory 500 stores a software logic 51, and the read-only memory 500 is connected to the microprocessor 3. The random access memory 400 is coupled to the microprocessor 3' to temporarily store the segment geometry coordinate data 230. During processing, the microprocessor 300 establishes a correlation between the segment feature point geometric coordinates 23G of the segments of the finger 210 according to the software logic 5, in the present invention, the layer is combined to contain a complete feature point. The complete feature point geometry of the moment $24G is 250. ^, 2 embodiment is different from the conventional hand-fingerprinting technology. When the surface of the sensor is 2 〇〇 sliding over the sensor surface, the captured slice & gray k-grain image is transmitted through the output logic 6 (10) Instantly the gray scale: the image is directly extracted from the feature point matrix, as shown by the label in FIG. 4, each output image is already converted to the segment feature before being temporarily stored in the random access memory 400. Point geometry coordinate data 230 (circle, pinch of the collection port), which is characterized by a relatively small amount of data for each segment feature point geometric coordinate data 230 (only tens of bytes of sand S even if the finger 200 slides over the sensing The surface of the device is extracted from 100 segments, and the text is converted into a segment feature point. The geometric coordinate data 230 is output and temporarily stored in the random access memory 400. It only needs a few thousand bytes (kbytes) of space. Figure 5 shows a schematic diagram of the architecture of the continuous fingerprint image capturing device according to the second embodiment of the present invention. Figure 5 and Figure 4 show the structure of the continuous fingerprint image capturing device according to the second embodiment of the present invention. Implement 13 1298852 examples of continuous fingerprint images撷The device comprises a sliding fingerprint sensing state 100, a processing state 300, a read-only memory 5, and a random access memory 400. The above-mentioned components and actuation modes are similar to the first embodiment, and only The difference from the first embodiment is described. U-processed 3 00 receives a plurality of segment fingerprint images 2 丨〇, and only software logic 51 〇 (4) is stored in the 呗 呗 memory 500. The microprocessor 300 is based on the software logic 51 〇 The segment fingerprint images 21〇 are extracted from the plurality of segment feature point matrices 22〇, and the segment feature point geometric coordinate data 230 is established according to the pieces of the slave point matrix 220 and temporarily stored in the random access memory. In 4(10), the pieces of feature points geometric coordinate data 230 of the pieces of the minions are then used to establish a correlation to synthesize the 70 feature point geometric coordinates of the complete feature point matrix 24 2 2 5 0. In the above two embodiments, the body logic 510 or the output logic 600 stored in the read-only memory 500 can provide the microprocessor 3 to perform instant feature extraction and geometric coordinates on the parent-segment fingerprint image. Capital, Operation logic. Since the gray-scale image is directly extracted from the feature coordinates in the embodiment of the present invention, % does not need a conventional processing procedure, such as gray-scale fingerprint image, image pre-processing and second-order, thin line, and finally The extraction is not required. Therefore, the complicated pre-processing and second-order processing are not required, so that the microprocessor 300 can be greatly saved, and the feature extraction method is quite simple: The hardware output logic 600 performs this function. The wheel-out logic itself does not need to have floating-point arithmetic functions, so it can be easily integrated into a single chip of the fingerprint sensor. 14 1298852 As shown in Fig. 5, each segment fingerprint image is immediately processed into a piece of slave feature point geometric coordinate data 23, and all are temporarily stored in the random access memory 400 towel. The software logic (aborism) in the read-only memory 5〇〇 also provides a geometric coordinate superposition operation logic of a segment feature point, and all the segment feature point geometric coordinate data 230 are superimposed into the complete feature point geometric coordinate data 250, as shown in the figure. The right side of 4 is shown. Fig. 6 is a view showing the private method of the method of the continuous fingerprint image capturing device of the present invention. As shown in FIG. 6, the processing method of the present invention includes the following steps: /S1 〇·Start, which may be to start fingerprint login or start comparison of buckles, in the case of starting to log in fingerprints, read only memory 5 (8) and 'There is a reference template stored, or the reference template stored in the read-only memory 5 will be cleared, and in the case of starting the comparison fingerprint, the reference template 500 already stores the reference template; Step S20· · capturing a plurality of segment fingerprint images 21 that are swiped through one of the finger 200 of a sliding fingerprint sensor ι; Step S30: Instantly extracting the segment fingerprint images 210 into a plurality of segment feature points a matrix 220; Step S40: establishing segment feature point geometric coordinate data 23〇 according to the segment feature point matrix 22〇 and temporarily storing it; Step S50: arranging the segment fingerprint image 21 to the segment feature point geometric coordinates #23〇 Establishing a correlation to overlay a complete feature point geometric coordinate data 250 containing a complete feature point matrix 240. Wherein, in the fingerprint login mode, the reference plate formed by the complete feature point geometric coordinate data 250 can be in the body of the read only memory 15 1298852; or the comparison 楹4 ΠΓ 1 can be completed, the work can be complete The feature point geometric coordinate poor material 250 is compared with the reference template stored in Weijing% "Bei" 丨 丨 丨, to confirm whether the user is an authorized user; and step S 6 0 ·· In the local area, ^ system, Ό bundled login or fingerprint comparison. The advantage of superimposing the feature point geometric data of the segment feature, in addition to the upper = can save the space of the random memory to reduce the system design, the complexity of the operation can be reduced, because the hand fingerprint road image in the prior art In the figure, it is necessary to compare the two fingerprint segments of the fingerprint image to make a full map comparison to find the minimum variation as the benchmark for the connection. The conventional method is the block comparison often used in image processing (bi〇ck she (4)). This method is computationally intensive and time consuming. However, in the method adopted by the embodiment of the present invention, only a few feature point geometric coordinates of the two geometric coordinate data need to be superposed logical operations, so that the calculation is small and time-saving. And the superposition of the geometric coordinates is quite easy for the rotation, ', if the situation occurs when the buckle is slid, the hand-fingerprint diagram of the prior art is difficult to solve because the second picture If there is an angle with the first image (for example, the sensor specification is 8*280, which is the Atmel Fingerchip specification), then the angle difference between the two is more than 1 degree, and the two fingerprint images are partially The areas cannot overlap, which makes the difficulty of the connection happen, which is a technical bottleneck that is difficult for the prior art to break through. However, for the embodiment of the present invention, the rotation of the feature point geometric coordinates can still easily complete the overlapping action, and will not Lost finger information. 16 1298852 Fingerprint caused by different speed and pressure when sliding with a stupid person's finger. II. Shape problem. The method of using the fingerprint road map is used. Every -, brother's and: Ϊ image The different pressures come from different use of the situation = the non-fixed behavior of the movement, so the joint fingerprint image 4 is extended or shortened, or in part 2: riding* - part of the region is a shortened situation occurs, and mutual authentication between the unfavorable performance of identification than or different sensor systems,
二如網路的兩端分別使用滑動式及面積型指紋感测 裔’則有可能無法相互比對。Second, if both ends of the network use sliding and area fingerprint sensing, they may not be able to compare each other.
在本貫施例中’每一片段指紋即時萃取出的複數 ㈣徵點幾何座標資料係不受變形所影響，原因為每 一片段指紋受滑動導致壓力變形的方向及大小是可以 視2相同的（因其擷取時間相當短，約幾毫秒），因此 所萃取出的複數特徵點間的向量係不變的，理所當然 的萃取出的複數個特徵點幾何座標資料的相對性:^ 便的，因此最後所疊合出來的完整特徵點幾何座標資 料與面積型指紋感測器所得到的是幾乎無異的。、 在較佳實施例之詳細說明中所提出之具體實施例 僅用以方便說明本發明之技術内纟，而非將本發明狹 義地限制於上述實施例，在不超出本發明之精神及以 下申請專利範圍之情況，所做之種種變化實施，皆屬 於本發明之範圍。 17 1298852 【圖式簡單說明】 圖所示為習知應用滑動式指紋感測器作 指紋圖像讀取的架構示意圖。 w曰 圖2所示即為利用圖1加 之木構所擷取到的連續降 片段指紋圖像以及疊合成之^ Μ α 』扪運、、，貝14 一 且口成之元整指紋圖像之示意圖。 圖3顯示依據本發明箆—每 弟貝知例之連績性指紋影 像擷取裝置的架構示意圖。 '圖4所不即為利用圖3之架構所擷取到的連續性 ?曰、文SU象即時片段特徵點幾何座標資料萃取以 且口成之兀整特徵點幾何座標資料之示意圖。 圖5顯示依據本發明第二實施例之連續性指紋影 像擷取裝置的架構示意圖。 圖6顯不本發明之連續性指紋影像擷取裝置之處 理方法之流程圖。 【主要元件符號說明】 S10至S60〜處理方法之步驟 1〇〜滑動式指紋感測器 20〜手指 21〜片段手指紋路圖像 22〜完整手指紋路圖像 3 〇〜微處理器 40〜隨機存取記憶體（RAM) 50〜唯讀記憶體（ROM) 100〜滑動式指紋感測器 2〇〇〜手指 18 1298852 210〜片段指紋圖像 220〜片段特徵點矩陣 23 0〜片段特徵點幾何座標資料 240〜完整特徵點矩陣 25 0〜完整特徵點幾何座標資料 300〜微處理器 400〜隨機存取記憶體（RAM) 500〜唯讀記憶體（ROM) 510〜軟體邏輯 600〜輸出邏輯In the present example, the geometric coordinates of the complex (four) sign of each segment fingerprint are not affected by the deformation, because the direction and size of the pressure deformation of each segment fingerprint can be regarded as the same. (Because its extraction time is quite short, about a few milliseconds), the vector between the complex feature points extracted is unchanged, and the relativeness of the geometric coordinates of the extracted multiple feature points is taken for granted: ^ Therefore, the final feature point geometric coordinates of the final overlay are almost identical to those obtained by the area fingerprint sensor. The detailed description of the preferred embodiments of the present invention is intended to be illustrative only, and not to limit the invention to the above embodiments, without departing from the spirit and scope of the invention The scope of the patent application and the various changes made are within the scope of the invention. 17 1298852 [Simple description of the diagram] The figure shows the schematic diagram of the conventional application of the sliding fingerprint sensor for fingerprint image reading. w曰 Figure 2 shows the continuous descending segment fingerprint image captured by the wood structure of Figure 1 and the composite image of the ^ Μ α 扪 、, , , , , , , , , , , , , , , , , Schematic diagram. Figure 3 is a block diagram showing the architecture of a collusion fingerprint image capture device in accordance with the present invention. Figure 4 is not a schematic diagram of the continuity of the structure of Figure 3 using the continuity of the structure of Figure 3, and the textual coordinates of the feature points of the instant segment feature points. Fig. 5 is a block diagram showing the construction of a continuous fingerprint image capturing device in accordance with a second embodiment of the present invention. Figure 6 is a flow chart showing the method of the continuous fingerprint image capturing device of the present invention. [Main component symbol description] Steps S10 to S60~ Processing method 1〇 Slide fingerprint sensor 20~ Finger 21~ Fragment hand fingerprint road image 22~ Complete hand fingerprint road image 3 〇~Microprocessor 40~ Random memory Memory (RAM) 50~Read Only Memory (ROM) 100~Sliding Fingerprint Sensor 2〇〇~Finger 18 1298852 210~Flip Fingerprint Image 220~Segment Feature Point Matrix 23 0~Segment Feature Point Geometric Coordinates Data 240 ~ complete feature point matrix 25 0 ~ complete feature point geometry coordinates data 300 ~ microprocessor 400 ~ random access memory (RAM) 500 ~ read only memory (ROM) 510 ~ software logic 600 ~ output logic