指紋辨識方法研究

The study of Fingerprint Identification

李泓汶(Hung-Wen Lee)

指導教授 ： 丁鏞

中原大學/工學院/機械工程研究所/碩士(2008年)

https://doi.org/10.6840/cycu200800142

摘要

摘 要

本研究之主要目的為設計一套應用於產業製程資料安全控管機制專用之即時指紋辨識系統。使用者初次登入系統時必須先進行指紋註冊登錄工作，再次登入後即透過指紋辨識進行身分認證。在指紋辨識系統中，分成三個單元。在指紋影像輸入單元中，利用RF射頻指紋感測器對手指真皮層進行指紋掃描，以提高指紋辨識的正確率。在指紋特徵處理單元中，透過影像前處理包括雜訊平滑、影像增強、細線化等程序來強化指紋影像；透過核心偵測包括流向計算、中心點判斷等程序來定義指紋中心；透過特徵擷取處理包括Gabor濾波器、扇形化、正規化、迴旋運算等程序來進行指紋特徵擷取。將所取得之指紋影像特徵值設定門檻值，並與使用者資訊一併存入資料庫中。在指紋識別比對單元方面，採用歐氏距離比對法(Euclidean distance)計算變異量與門檻值進行比對，驗證成功則允許使用者登入，反之則彈出警告訊息並要求使用者重新輸入指紋。在模擬階段運用倒傳遞類神經網路(BPNN)來進行資料訓練，以提高整體辨識率。本研究設計之實作系統，其執行驗證所需時間小於4.5秒，錯誤接收率(FAR)小於0.003%，錯誤拒絕率(FRR)小於0.03%，由於系統辨識速度快，正確率高，適合建置於使用者眾多的業界環境使用。

關鍵字：指紋辨識，特徵擷取，倒傳遞類神經網路。

關鍵字

倒傳遞類神經網路 ； 指紋辨識 ； 特徵擷取

並列摘要

ABSTRACT

The objective of this study is to design a real-time Fingerprint Identification system for the industrial process data security control. Users must complete the fingerprint registry steps and acquire authorization of the fingerprint identification system so as to login.
There are three primary units in this system. In Fingerprint Input System, RF sensor is used to scan fingers in order to improve the accuracy. In Fingerprint Feature Process System, image is pre-processed by using image smooth, image enhancement, and image binarization so that image quality is enhanced. Through the process of Fingerprint Direction Computation and Core Point Detection, core point is defined. In the process of Feature Extraction including Gabor Filter, Fan-shape, Normalization and Convolution, features are extracted easily. When capturing features, threshold of each fingerprint is designed and saved into the database system.
The Fingerprint Identification Process System is designed based on Euclidean distance algorithm to calculate and indentify the features between the new fingerprints and sample templates. If identification is passed, system will automatic carry out user information; on the contrary, it will show error messages and ask users to input again. Propagation Neural Network (BPNN) for data training is used to increase accuracy of the Fingerprint Identification. In this study, the system execution time measured is less than 4.5 seconds, and the False Accept Rate (FAR) is less than 0.003%, and the False Reject Rate (FRR) less than 0.03%. With fast identification speed and high accuracy rate, the developed identification system is suitable for various application purposes in industry.

Key words：Fingerprint Identification, Feature Extraction, Back Propagation Neural Network (BPNN).

並列關鍵字

Back Propagation Neural Network (BPNN) ； Feature Extraction ； Fingerprint Identification

參考文獻

[3] E. R. Henry, Classification and Uses of Finger Prints, Routledge, London, 1900.

[4] M. Kawagoe and A. Tojo, “Fingerprint pattern classification,” Pattern Recognition, vol. 17, No. 3, pp. 295-303, 1984.

[6] B. M. Mehtre, N. N. Murthy and S. Kapoor, “Segmentation of Fingerprint Image Using The Directional Image” Pattern Recognition, vol. 20, pp. 429-435, No.4, 1987.

[7] A. Grasselli, “On the automatic classification of fingerprints some consideration on thelinguistic interpretation of pictures,” Methodologies of Pattern Recognition, pp. 253-273.

[8] B. Moayer and K. S. Fu, “An application of stochastic languages to fingerprint patternrecognition,” Pattern Recognition, vol. 8, pp. 173-179, 1976.