【摘要】 由于指纹的匹配是根据指纹特征点匹配的,所以指纹特征点提取方法的好坏将直接影响指纹匹配的成功率。针对指纹特征点提取耗时较长的问题,提出了一种在细化指纹上,利用支持向量机的特征点(端点及分叉点)提取方法,首先进行指纹特征点初提取,即将特征点的8邻域模板作为输入,特征点类型作为输出,建立支持向量机网络并进行训练。将任意的纹脊点的8邻域输入到已训练好的支持向量机网络就可得到特征点的类型。初提取得到的指纹特征点集中存在大量伪特征点,判断图像边缘并根据特征点之间的方向及距离消除伪特征点。仿真结果表明,改进方法指纹提取的速度快,同时去除伪特征点的准确率高。更多还原

【Abstract】 Fingerprint matching is often based on minutiae matching,so quality of minutiae extraction method affects the success rate of fingerprint matching directly. For the fingerprint feature points extracted takes long time,this paper introduces a method that extracts the minutiaes( endpoints and branch point) on fingerprint which have been refined based on support vector machine( SVM). This method takes 8 neighborhood points of minutiae templates as input,feature points type as output. Then a support vector machine( SVM) network is established and trained. The 8neighborhood of ridge points are inputed into the well- trained support vector machine( SVM) network,and the false minutiaes are eliminated according to the direction and distance of the points,then we can get the type of the point.Simulations show that this method is fast to extract the minutiaes,and it can remove false minutiaes correctly.更多还原