中文摘要：

针对传统的奇异点检测方法主要基于方向场变化且容易受噪声影响，提出了一种新的基于模型分析的指纹奇异点检测方法．首先用基于离散余弦变换基函数对方向场进行建模，在计算方向场的基础上。利用常微分方程系统线性化数学模型，通过对模型参数和平衡点进行分析，检测指纹奇异点位置．实验结果表明，基于常微分方程线性化模型分析的指纹奇异点检测方法比传统的Poincare Index方法对噪声更具有较好的鲁棒性，能进一步提高奇异点检测的准确度．

英文摘要：

As traditional singular point detection methods are mainly based on the orientation changes which are oflen affected by noise, a novel method was proposed for fingerprint singular-point detection based on mathematical model analysis in this paper. First, discrete cosine transform （I）CT） was used as lhe basis functions to build the orientation reconstruction model and compute the orientation field. Then, the ordinary differential equation and system linearization model were computed with fingerprint orienta tion field and the singular points detected based on the model analysis. Experimental resuhs show that the proposed singular point detection method has high robustness to noise and can achieve a higher accuracy compared to the traditional Poincare Index method.