开放式环境非接触低分辨率采集下的掌纹识别

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.     

An Automatic Digital Camera Based Fingerprint Verification System

The emergence of touch-less fingerprint recognition system brings both advantages and challenges in the automated fingerprint recognition research. In this paper, we proposed a touch-less fingerprint recognition system by using a digital camera. The system frees from the problems in term of deformation caused by the contact pressure, hygienic, maintenance and latent fingerprints that exist in the touch-based fingerprint recognition system. However, fingerprint images that were acquired with digital camera consist of low ridge-valley contrast, defocus and motion blurriness problems. Hence, preprocessing steps are introduced to reduce those problems. The system comprises of preprocessing, Gabor filter feature extraction, Principle Component Analysis and verification using Support Vector Machine. Satisfactory results are attained in our acquired fingerprint database.