基于细节特征点的掌纹比对算法及GPU加速

Palmprint matching algorithm based on minutia feature points and GPU application

针对当前公安机关在掌纹识别应用中存在的掌纹库容量大和现场掌纹质量差等问题,首先对掌纹特征点匹配算法的计算复杂度进行分析,根据分析结果提出特征点的三级比对算法;然后详细介绍三级比对算法的工作流程;对算法进行GPU并行化改造;最后给出测试结果.首次将GPU计算技术应用于掌纹识别之中.文中掌纹特征比对方法相比现有同类掌纹识别速度提高3倍左右,经过GPU加速后可进一步提升比对速度15倍以上.

The problems of palmprint application in public security department,such as large database and poor quality of scene palmprint,are described in this paper. Firstly,the computational complexity of palmprint feature point matching algorithm is analyzed. The three-level feature point matching algorithm is presented based on the results of the analysis. Secondly,the working process of three-level feature point matching algorithm is described in detail. Finally the algorithm is transformed to the GPU parallel and the test results are given. The GPU technology is first applied to palmprint recognition in our study. The results show that the speed of palmprint feature matching increases about three times compared with others,and the speed may increase more than 15 times after using GPU technology.