多级提取奇异点的算法及基于奇异点的粗匹配

提出了一种分别基于Poincare指数和梯度场零点检测的分级算法来提取奇异点(核心点,三角点).该算法分为两级:首先是粗定位,根据指纹块方向场的Pioncare指数确定奇异点所在的分块和该奇异点所属的类型,同时得到核心点的方向;然后精定位,用梯度场零点检测的方法在该块中精确确定奇异点的位置(精确到像素级);最后利用奇异点的位置和方向信息对指纹做粗匹配,尽可能地排除明显不可能的候选待匹配指纹.实验证明,提取奇异点的算法快速、准确并且有很强的鲁棒性,而且与全图平方方向场滤波算法相比节省27%的计算时间.后阶段粗匹配的算法则能够拒绝30%的候选指纹系统的处理时间.

Algorithm of multi-stage detecting singular point and coarse matching based on singular points

On the basis of Poincare Index and zero point-detection of gradient field,a hierarchy algorithm for detecting singular point(core and delta) was proposed.It was divided into two stages.One was a coarse location stage: according to Pioncare index of fingerprints block orientation field,the block in which singular points(SPs) located and the types that SPs belonged to were defined.The direction of core was obtained at the same time.The other stage was precision location.The location of SPs was determined in this block-precise-pixel-level by zero point-detection of gradient field.At last,a coarse matching algorithm was used for fingerprint by means of SPs direction and direction information,eliminating obvious impossible waiting-for-matched fingerprints.The experiment proved that the algorithm with strong robust for detecting SPs was rapid,accurate,and of strong robust.Compared with the algorithm for filtering the orientation field in an image,this algorithm can save 27?% of the computation time.In the later stage,30?% of waiting-for-matched fingerprints could be rejected and the processing time of the system be saved in this algorithm.