基于细节特征的自动指纹识别系统的研究

【作者】 曹祥；

【导师】 邓宏贵；

【作者基本信息】 中南大学 ， 光学， 2007， 硕士

【摘要】 近几年，指纹识别技术获得了较大的发展，但仍是目前研究的热点，主要集中在如何提高指纹识别系统的准确率和识别速度，因此无论是在理论上还是在应用上研究自动指纹识别技术都具有非常重要的意义。在自动指纹识别系统中，提取到的指纹特征的质量对指纹特征匹配有很大的影响。对于高质量的特征，目前的算法都能得到一个很好的匹配结果。此外，在指纹识别系统中，指纹特征的提取耗时最多。所以对指纹特征提取的研究是当前和今后指纹识别研究的重点。本文通过对指纹特征提取全过程的系统分析，对其中指纹图像的分割、纹线方向计算、纹线周期的计算、图像的增强、细节特征提取和特征匹配提出了一些新的见解。主要包括：改进了基于梯度的方向计算方法；提出了基于特征融合的指纹图像分割算法；提出了基于一维傅立叶变换的纹线周期提取算法；提出了新的Gabor滤波函数参数的设计方法，提高了滤波器的自适应性；在细节特征的提取过程中，设计了基于方向和纹线波形的二值化方法，引入了新的细节特征滤除算法，避免毛刺消除和断线连接步骤带来的时间开销，提高了细节特征的质量；在指纹特征匹配部分，设计基于统计的平移旋转参数的确定方法，克服了目前基于细节特征匹配方法的缺点。分析研究表明，新的指纹图像预处理算法从整体上提高了系统的处理效果，降低了时间复杂度；基于统计的细节特征匹配算法提高了匹配算法的鲁棒性，降低了错误拒绝的概率。更多还原

【Abstract】 In recent years, there is a great progress in fingerprint identificationtechnology, but it still is a hot topic at present, which is focused on how toimprove the accuracy and recognition speed of the automatic fingerprintidentification system. Therefore, it’s very important to research the automaticfingerprint identification technology both in theory and in application.In the automatic fingerprint identification system, the fingerprintmatching relies on the quality of the features heavily. When the quality of thefeatures is very well, all the algorithms can get a good result. In addition,fingerprint feature extraction process is the most time-consuming in theautomatic fingerprint identification system. So the fingerprint featureextraction is the current and future key problem in fingerprint identificationtechnology. By analysis of the fingerprint feature extraction process whichincludes the fingerprint image segmentation, computing ridge direction,computing ridge period, image enhancement, minutiae extraction and detailsof the matching, some new views are proposed, which mainly include:improving orientation extraction algorithm based on the gradient, based onthe integration of the fingerprint image segmentation algorithm, based on theone-dimensional Fourier Transform ridge period computing algorithm, a newway of Gabor filter function designed which improves the adaptive capacityof the filter. In the minutiae extraction process, based on the direction andridge waveform values of the binary method is designed, a new featuredetails filtering algorithm is introduced which can retrench the resuming timeof eliminating burr and improve the reliability of the details. In thefingerprint matching process, A statistic model is proposed, which cancompute the move-rotation parameters not from the local of fingerprint butfrom the whole fingerprint, overcome the shortcomings of the presentminutiae-matching method. Analysis and some experiments show that thesenew algorithms overcome some shortcomings of present algorithms, improve the performance of the Automatic Fingerprint Identification System. Thematching algorithm based on the statistical enhances the robustness, reducesthe error probability of rejection.更多还原