低價格與小體積之指紋辨識系統

The Low Cost and Small Size Fingerprint Verification System Base on Line Sensor

余信緯(Hsin-Wei Yu)

指導教授 ： 黃文增

國立臺北科技大學/機電學院/電腦與通訊研究所/碩士(2006年)

https://doi.org/10.6841/NTUT.2006.00222

摘要

科技進步與其產品多元化，享受科技產品便捷之時，個人資料的安全性就突顯重要。在傳統上，大多電子相關產品乃是以輸入密碼方式當成安全性的防護鎖，但密碼較易於遭人竊取與記憶不便等缺點。在本文中，研究與實作條狀感測器所建構成的指紋辨識系統作為安全防護；所使用的指紋條狀感測器比傳統指紋塊狀感測器在特性上有低功率、小面積、與低價格的優勢。
條狀感測器是使用刷取方式來擷取片段指紋，將已經擷取到的片段影像經過重組成一個完整的指紋影像。因為在前端指紋刷取感測器時，有刷取速度快慢、指壓不同、和工作環境變數等問題；因此，必須應用後端的補償演算法來加以修正與校調處理，以避免這些問題造成影響辨識結果。
首先，本文使用正規化交相關重組演算法來重組片段指紋成為一個完整指紋。然後，將完整的指紋影像做增強功能，如柱狀圖平均法、高斯濾波、和圖像縮放，方便取其特徵值。使用群延遲頻譜演算法去擷取影像增強後的“指紋頻譜特徵”值。最後，使用動態時間規劃演算法將指紋頻譜特徵值和已經註冊的指紋作比對驗證，以完成辨識步驟。
因為，本系統採用群延遲頻譜的頻譜特徵值取樣，和傳統所使用的端點與叉點特徵值相比，所需的影像前處理計算量相對地少與較少的影像增強功能；因此，本系統有低計算量和需求較少記憶體的優勢。本系統基於條狀感測器和群延遲頻譜有著上述的優勢；更進一步，經由實驗得知本論文的辨識率為 93.82%。

關鍵字

條狀指紋感測器 ； 正規化交相關法 ； 群延遲頻譜 ； DP比對 ； 閥值

並列摘要

While the diversity of the technology has been presented and advanced nowadays, it is time to enjoy the convenience from the progress of technology. Thus, the way to secure the personal data has been a more and more important issue. Traditionally, products related to electronics adopt the password inputting as the protection lock. But, the password is inconvenient since it acquires the user to memorize and it can easily be stolen. In the present study, a fingerprint verification system is the main focus in which the line sensor has been adopted to be the protection lock. The line sensor used here, requiring lower power, smaller area, and lower price, is different from the traditional block sensor, for the current system of science and technology on the market. This kind of sensor will be much suited for the product development.
Because of the line sensor, the way to detect is to sweep and pick up the fingerprint, and, thus, the image that have already picked can be further reorganized to become as a full fingerprint image. Because the head fingerprint sweep on the sensor, having problems such as sweep speed, pressure differences, and work environment variables. The algorithm of backend expiation has to be applied to take for a further revision and for the adjustment of processing, to avoid problems that influence the recognized result.
First, this present study adopts the Normalized Cross Correlation reorganization algorithm to recognize the complete fingerprint. Then, in order to enhance the function of a complete fingerprint image establishment, as the Histogram equalization, Gaussian smoothing, and image scaling, it is convenient to take its minutiae. We use the group delay spectrum algorithm to extract the fingerprint spectrum minutiae after the image has been enhanced. Finally, we use the dynamic programming algorithm to match and verify between fingerprint spectrum minutiae and the registered fingerprint, to complete the recognition.
Because this system uses the spectrum minutiae of the group delay spectrum to be as the sample, it can be used to compare with the traditional endpoint and bifurcation minutiae. Since this system contains the advantage with lower calculation quantity and less memory, it requires much less calculation for the quantity of image front processing and image enhancement function. This system is based on a line sensor and group delay spectrum in which the relevant advantages have been mentioned above. Therefore, through the experiment, our recognition rate is up 93.82%.

並列關鍵字

Line sensor ； Normalized Cross Correlation ； Group Delay Spectrum ； DP matching ； Threshold

參考文獻

[24] 曾銘崧，熱感式指紋成像及其自動辨識法之研究，碩士論文，私立中原大學電子工程系，桃園，2001。

[3] N. Ratha, R. Bolle, Automatic Fingerprint Recognition Systems, NewYork:

[7] K. Briechle, U.D. Hanebeck, "Self-localization of a mobile robot using fast normalized cross correlation," IEEE SMC '99 Conference, Tokyo, Japan, 1999, pp. 720-725.

[8] B.H. Friemel, L.N. Boho, G.E. Trahey, "Relative performance of two-dimensional speckle-tracking techniques: normalized correlation, non-normalized correlation and sum-absolute-difference," Proceedings of 1995 IEEE Ultrasonics Symposium, vol.2, 1995, pp. 1481-1484.

[13] N. Matsumoto, S. Sato, H. Fujiyoshi, and T. Umezaki, “Evaluation of a Fingerprint Verification Method Based on LPC Analysis,” T. IEE Japan, Vol.122-C, No.5, 2002-5, pp.799–807(in Japanese).