开放式环境非接触低分辨率采集下的掌纹识别  

Touch-less low resolution palmprint recognition under open environment

作　　者：桑海峰[1] 马诗朦[1] 黄静[1] 

机构地区：[1]沈阳工业大学信息科学与工程学院沈,阳110870

出　　处：《仪器仪表学报》2013年第8期1802-1808,共7页Chinese Journal of Scientific Instrument

基　　金：高等学校博士学科点专项科研基金(20122102120004);辽宁省教育厅科研基金(L2012034)资助项目

摘　　要：提出一种开放式环境下、非接触采集下的掌纹识别算法。采用一种低分辨率的普通摄像头进行掌纹图像采集,人手只需自然张开、中指竖直向上放在摄像头前一定范围内。首先建立肤色模型将人手从复杂背景中提取出来;然后提出一种简单有效的关键点定位算法找到指根点,并以此建立坐标系提取掌纹ROI;接着提出一种基于统计的纹理基元算法进行掌纹特征提取,该算法不用对掌纹ROI进行归一化,计算速度较快并且对光照具有一定的鲁棒性;最后利用余弦相似度进行特征匹配实现掌纹识别。在自建的小型掌纹图像库上进行了相应的识别实验,实验结果表明所提出的算法能够在动态环境下达到理想的识别效果。所开发的实验系统能够快速实现掌纹图像采集与识别,适合在线实时系统应用。This paper proposes a touch-less palmprint recognition algorithm in open environment. The system uses a low-resolution camera to achieve the palmprint image acquisition. In the acquisition, the hand is required to naturally open, and the middle finger is put straight up in front of the camera within a certain range. First, the skin-color model is established to extract the human hand from the complex backgroun tioning algorithm is proposed to find the finger root points. Based on lished to extract the region of interest （ROI） of the palmprint. After d. Second, a simple and effective key point posi- these key points, a coordinate system is estab- that, a new algorithm that applies texture primi- tive based on statistics is proposed to extract the palmprint features. This algorithm does not require aligning all the ROIs to the same location and has the advantages of fast computing speed and robustness to the light illustration. Fi- nally, the cosine similarity algorithm is deployed to perform feature matching and palmprint recognition is achieved. The proposed algorithm was tested in the recognition experiment on a self-built database. Experiment results show that the proposed algorithm achieves good results under dynamic environment;the developed system can quickly realize the acquisition and recognition of the palmprint image, and is suitable for online real-time system applications. ：

关 键 词：掌纹识别 非接触 纹理基元 复杂背景 在线识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]