臺灣博碩士論文加值系統

長久以來指紋已被廣泛用於作為生物辨識技術的一個重要指標，其主要原因在於每一枚指紋皆不相同且終生不變之特性，單一指紋的採集即使受到設備或演算法的影響，對於指紋辨識的失誤率能只有百萬分之一之小，相較於其他種類的生物辨識技術而言仍是便是精準度最高的技術，在過去刑事偵查上也都作為犯罪現場證物蒐集的重大證據之一。但犯罪現場所採集的指紋經常受到不同的因素所污染，其中之一便是多枚指紋重疊時，所造成指紋辨識度降低，而需要利用大量人力將重疊指紋分離。 因此本論文研究提出一個方法，透過研究多種特徵對於紋理特徵擷取的方法，找出一種可以快速分離出重疊指紋影像區域，並分析出重疊指紋影像中各指紋紋理的方向場，再於小區域將重疊指紋影像分離，找出特定的特徵點，快去比對重疊指紋影像裡是否有無出現在單一指紋影像中。

In this thesis, two key techniques, the multiple filter set and the feature classification method based on support vector machines (SVMs) are proposed for fingerprint feature identification. In the past approaches, the Gabor filter was verified as a successful method, and usually selected as fingerprint feature detection. Although the Gabor filer have high accuracy for feature extraction under widely angle representation range, not all degree ranges of the filter are necessary. A disadvantage in the different degree range will cause large computation costs. Moreover, fingerprints have clean direction field representation which can be used to design suitable filter set for low computational complexity. Due to two kay factors, this thesis adopts the modified Haar-like patterns to perform near-circle filter set for acceptable feature coverage. In the processing of overlapping fingerprint separation, this thesis also proposes an efficient method based on statistic results of frequency domain response under Discrete Wavelet Transform (DWT). In feature identification, the efficient SVM-based classifier are presented. For more widely comparison, two kernels of linear and radial basis function were selected and implemented. In performance comparison, using general fifteen Haar-like pattern as benchmark, the proposed algorithm can reduce about 50% computation in average and keep on equal accuracy. Based on accuracy and computation complexity, our proposed can reach exactly real application.

摘 要 i ABSTRACT ii 誌 謝 iv 目 錄 v 表 目 錄 vii 圖 目 錄 viii 第一章 緒論 1 1.1 研究背景與動機 1 1.2 研究目的 2 1.3 論文架構 2 1.4 本文貢獻 2 第二章 文獻回顧 4 2.1 指紋概論 4 2.1.1 指紋紋理 4 2.1.2 指紋特徵點描述 5 2.2 紋理特徵的分析 6 2.2.1 二維傅立葉變換2D-DFT 6 2.4 現有重疊指紋分離文獻 9 第三章 研究方法 12 3.1 重疊指紋偵測與分割 13 3.1.1指紋影像之定義 13 3.1.2 離散小波轉換特徵擷取 14 3.1.3 指紋紋理種類分析實驗 17 3.1.4 指紋影像分割 19 3.2 重疊指紋方向場分析 20 3.2.1 影像周邊雜訊濾除 20 3.2.2 重疊指紋頻域信號變換 21 3.2.3 重疊指紋頻域信號前處理 22 3.2.4 改良式重疊指紋方向場分析 25 3.3 快速重疊指紋分離 27 3.3.1 頻域信號之濾波 27 3.3.2 頻域遮罩產生及濾波 28 3.3.3 指紋頻譜信號反離散傅立葉轉換 29 3.4 指紋特徵點匹配 30 3.4.1 指紋影像細化與特徵點擷取 31 第四章 實驗結果與分析 33 4.1 重疊指紋影像切割結果 33 4.2 重疊指紋分離結果 34 4.3 指紋特徵點偵測結果 37 4.3.1 指紋特徵點比較 38 第五章 結論與未來工作 39 5.1 本論文結論 39 5.2 未來發展 39 參考文獻 41

[1] The Statistics Portal-http://www.statista.com/statistics/194213/crime-clearance-rate-by-type-in-the-us/ [2] Open Cases: Why One-Third Of Murders In America Go Unresolved-http://www.npr.org/2015/03/30/395069137/open-cases-why-one-third-of-murders-in-america-go-unresolved [3] Jianjiang Feng, Yuan Shi, and Jie Zhou, “Robust and Ef?cient Algorithms for Separating Latent Overlapped Fingerprints,” IEEE Trans. Inf. Foren. Secur., vol. 7, no. 5, pp. 1498-1510, 2012. [4] Qijun Zhao and Anil K. Jain, “Model Based Separation of Overlapping Latent Fingerprints,” IEEE Trans. Inf. Foren. Secur., vol. 7, no. 3, pp. 904-918, 2012. [5] 台灣刑事局https://www.cib.gov.tw/Science/EncyclopediaDetail/233 [6] L. Hong, Y. Wan, and A. K. Jain, “Fingerprint image enhancement: Algorithm and performance evaluation,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 20, no. 8, pp. 777–789, Aug. 1998 [7] Ruye Wang, Fourier Analysis http://fourier.eng.hmc.edu/e101/lectures/Fourier_Analysis_brief/Fourier_Analysis_brief.html [8] Rafael C. Gonzalez & Richard E. Woods(1997), Digital Image Processing [9] P. Viola and M. Jones, "Rapid object detection using a boosted cascade of simple features," 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, Vol 1, Proceedings, pp. 511-518, 2001. [10] A. Rosenfeld, R. A. Hummel, and S. W. Zucker, “Scene labeling by relaxation operations,”IEEETrans.Syst.,Man,Cybern., vol. SMC-6, no. 6, pp. 420–433, Jun. 1976. [11] Yi-Pin Hsu , Yen-Lin Chen, “Fast Fingerprint Feature Extraction Based on Modified HaarLike Patterns Using Support Vector Machine ’’., 2017 IEEE International Conference on Consumer Electronics - Taiwan (ICCE-TW) [12] 蔡鴻旭、顏志遠、侯能如，"基於支援向量機與小波影像紋理檢測瑕疵塑膠編織袋," 2010 資訊科技國際研討會(AIT2010), 朝陽科技大學, April 23-24, 2010. [13] 周建忠、夏榮生，「小波轉換於SAR影像資料萃取之研究」，航測及遙測學刊，2001，第六卷，第二期，第71-86頁 [14] 張業承，指紋影像清晰化的研究，碩士論文，國立交通大學資訊科學系，2004年 [15] 莊智瑋、林昭遠、陳正湘，「利用紋理因子改善影像分類準確度之研究」，水土保持學報，41，第153-168頁，2009 [16] 田文敏、郭孟維，「應用灰度伴隨矩陣進行側掃聲納之結構分析與海床分類」，第二十八屆海洋工程研討會論文集，pp. 759-764，2006