基于奇异点区域方向场的指纹检索

Fingerprint Indexing Based on Singular Points’ Orientation Field

为提高大型指纹数据库中指纹识别的速度和准确性,必须对其进行有效分类和快速检索。提出了一种基于奇异点区域方向场的指纹检索方法。首先利用支持向量机对指纹图像进行分割,并对分割出的图像的方向场进行多尺度平滑,得到可靠的平方复数点方向场估计,再对其进行复数滤波,利用滤波响应幅度信息确定奇异区范围及奇异点的位置和方向;然后根据所得奇异点个数及相对位置对指纹进行初次分类,最后利用奇异点区域的方向场构成指纹的特征向量,并通过比对特征向量进行指纹检索。实验结果表明,本文方法较文献的方法有明显的优势,能有效缩小待匹配指纹的数量。

A fingerprint indexing method based on singular points’ orientation is presented in order to accelerate the speed and improve the accuracy of fingerprint identification in the large-scale fingerprint database. Firstly, fingerprint image is segmented by using Support Vector Machine, and the square -complex orientation field is estimated through multiple scales smoothing in the segmented image. Secondly, the orientation field is filtered by a complex filter, and the filtering response amplitudes are obtained,which are used to determine the singular points＇ range, position and orientation. Finally, the first classification is made according to the singu- lar points＇ number and their relative positions. Fingerprint indexing is implemented through contrasting feature vectors which are composed of the orientation field around the singular points. The experiment shows that the method is better than the algorithms de- scribed in the references, which can effectively reduce the number of fingerprints to be matched.