基于Fourier-Mellin变换的指纹匹配算法研究

Research on Fingerprint Matching Using Fourier-Mellin Transform

为了克服指纹识别中平移和旋转等因素的影响,提出一种基于指纹整体信息相关匹配的算法。首先对待识指纹和模版指纹进行Fourier-mellin变换(FMT),然后提取FMT结果中的相位信息,运用带限相位相关(BLPOC)法得到待识指纹与模版指纹的相似度。因为FMT具有旋转特性,这样避免了传统的相关算法必须对每一个角度都进行运算的问题。同时与纯相位相关法相比(POC),在判断是否匹配成功时带限相位相关(BLPOC)法可以有效的抑制噪声并能获得更为陡峭的相关峰。通过计算机仿真结果表明算法指纹误识率和拒识率分别达到和,并在实际应用中取得了很好的效果。

To overcome the influence caused by shift and rotation in fingerprint matching, an correlation- based algorithm is presented. Fourier - Mellin Transform （FMT） is employed here, which can determine the rotation angle between the template and query fingerprint directly, while traditional correlation methods have to calculate the correlation over all possible rotations. The phase information is extracted from the result of FMT,and BLPOC is used to obtain the correlation of the template and query fingerprint images. The Band - Limited Phase - Only Correlation （ BL- POC） method is more robust to the noise and can provide a sharper peak to distinguish the genuine and impostor matching than the original Phase - Only Correlation （POC）. Learning from the result of simulation , EER is and FAR is.