一种自动指纹粗分类算法的研究

提出了一种利用指纹图像奇异点的数目和位置进行粗分类的算法。指纹图像大致可分为拱形、尖拱形、左旋形、右旋形和螺纹形5种，对指纹的粗分类是实现指纹自动识别系统的一个重要步骤。首先介绍了一种计算指纹图像中各个像素点的方向矢量的算法，然后详细讨论了在这个算法所获得的指纹方向图上进行平滑处理、提取奇异点对（核心点－三角点对）的算法，并根据奇异点对的数目和位置进行指纹方向图的粗分类。试验结果表明，有用的算法与Kalle的算法相比，具有较高的分类率，具有实用意义。

Study of an Automatic Fingerprint Coarse Classification Algorithm

In this paper, a fingerprint coarse classification algorithm based on the count and location of singular points in the Automatic Fingerprint Identification System (AFIS) is put forward. Fingerprints are normally classified into five categories: arch, tended arch, left loop, right loop and whorl. Fingerprint coarse classification is a very important step to realize the AFIS. Firstly, an algorithm of computing the directional vector of each pixel in a fingerprint image is introduced. Then, the smoothing algorithm and the extracting singular points (core-delta pair) algorithm upon the fingerprint directional image based on the above algorithm is discussed, and finally the fingerprint according to the number and the location of those singular points is classified. The experiment results show that the algorithm can achieve a higher rate of classification than Kalle's algorithm.