摘 要 指纹是当今应用最广泛的生物特征，指纹识别在人的身份鉴定中有着重要的作用，目前，大多数指纹识别平台都是连接PC机的桌面应用系统，这就导致了指纹识别系统的成本偏高，其应用得不到普及，因此，开发出识别率高，处理速度快、扩展性好、廉价的嵌入式平台有着广阔的市场前景和研究价值。 论文首先采用三星公司ARM9核心的S3C2440A处理器组建了嵌入式硬件结构，进行了外围电路的设计，并设计了一款实用的BIOS。在此平台上，完成了该款新型指纹传感器FPC1011F在ARM9系统的硬件和软件设计，并在MATLAB中将采集到的指纹数据进行了还原，证明了采集性能优良。 在采集到指纹的基础上进行了如下研究： 1.分别研究了灰度分割算法，Gabor滤波，固定阈值法，数学形态学细化算法，在VC++6.0平台上利用以上算法对指纹进行了分割、滤波增强、二值化、细化的指纹预处理。 2.在指纹预处理的基础上，研究了指纹提取算法，利用八邻域图法提取了指纹端点和叉点，并提出了利用方向场和Poincare公式结合的方法，对指纹奇异点的提取，并在VC++6.0平台上进行了验证。 3.在指纹特征点提取之后，利用本文中方向场和Poincare公式结合的方法提取的奇异点，提出了利用奇异点进行的校准方法，克服了利用脊线校准方法占用很大的内存空间和处理时间以及对质量较差指纹识别率低的缺点。 4.参考罗西平等人的算法，提出了一种基于极坐标转换，采用大小可变的限界盒来适应指纹的非线性形变的中心点匹配算法，进行指纹的匹配，利用FPC1011F指纹传感器采集的指纹图像数据库，在VC++6.0平台上按照测试标准验证了该算法。 经一系列实际采集识别实验证明，系统识别效果良好，证明了本文提出的匹配算法对质量不好的指纹能更好的识别，也能更鲁棒地处理指纹图像的非线性形变，降低了指纹匹配的误判率，缩短了匹配时间。为嵌入式指纹识别算法的移植进行了实验验证，为嵌入式识别系统的广泛应用打下了基础。

ABSTRACT Nowadays, the fingerprint is most widely used. Fingerprint biometric identification plays an important role in identity, at present, most of fingerprint identification platforms connect PC desktop application system, which lead the application of fingerprint identification system to the higher cost and not universal. Therefore, for the broad market prospect and the research value, it is necessary to develop embedded platform of high recognition rate, processing speed, good scalability and cheap. Firstly, designing a practical BIOS on the basis of embedded hardware structure forming with Samsung ARM9 processor core S3C2440A and the peripheral circuit designing. In this platform, the new type of fingerprint sensor in FPC1011F ARM9 system hardware and software designing both collecting in MATLAB fingerprint data, proving acquisition excellent performance. On the basis of acquisition to fingerprint as follows: 1. Using the VC 6.0 platform to research gray image segmentation algorithm, Gabor filtering, fixed threshold value method, mathematical morphology, refinement algorithm, and then research in advanced algorithms, filter enhancement, partition binary, refined and fingerprint pretreatment. 2. Researching the fingerprint extraction algorithm on the basis of the fingerprint pretreatment, using eight neighborhood figure law to extract a fingerprint endpoint and fork points, and propose using direction field and Poincare formula of the method of combining, the extraction of the fingerprint singularity, and verifying on the base of VC 6.0 platform. 3. After extracting the feature points of fingerprints, then using singular point calibration which extracted from singularity and Poincare form direction field, overcoming the shortcomings of great memory space because of using ridge line calibration method, processing time and poor quality fingerprint low recognition rate. 4. The algorithm of reference Rossi equality, which proposed based on polar conversion, adopting size variable confines the box to adapt the fingerprint nonlinear deformation, center matching algorithm, then verifying the algorithm the VC 6.0 platform in the test standard results with the FPC1011F fingerprint match fingerprint sensor acquisition fingerprint image database. Through a series of practical acquisition recognition experiment proof, it is effective of the identification system, which proves poor quality fingerprints can be better recognized on the matching algorithm in this paper, and also can be more robust to deal with the nonlinear deformation fingerprint image, reducing the rate of fake fingerprint matching and shorten matching time. The research laid the foundation for embedded fingerprint identification algorithm transplant validated by experiment and the wide application of the embedded identification system.