活体虹膜检测技术研究
发布时间：2019-05-23 16:40
【摘要】：随着近些年人工智能与图像识别领域呈现出的越来越火热的趋势以及智能电子设备数量和普及率的爆发式增长,让生物识别技术真正的融入到了人们的生活之中。虹膜识别技术快速发展的三十年来,国内外专家学者经过不懈的努力,已经提出并完善了许多经典的算法。但是无论识别技术如何先进,如果不能对伪造攻击手段做出有效的防御也不能被称作是合格的识别技术。本文主要针对已有基于瞳孔对光反射特性的活体检测算法的所存在的局限性结合了虹膜纹理变化检测共同组成分类特征来对相应攻击模型进行防御。并根据该算法无法很好防御的移动智能设备等新生攻击手段提出了基于双红外波段的活体虹膜检测算法。丰富完善了对于不同攻击手段的防御方法。本文的创新性工作可以概括如下:(1)提出了一种结合虹膜纹理和瞳孔反射特性的活体检测算法,在原有算法通过对瞳孔光反射特性检测的基础上结合对虹膜纹理特征的检测来防御通过位移变化等方法模拟瞳孔收缩变化的模型攻击。在该算法中为了对瞳孔部分更准确的进行分割提出了步进式自适应阈值选取算法,相比固定阈值的分割具有更好的鲁棒性,为后续的定位精准度提供保障。提出了改进的基于Hough变换的定位算法在原有基础上结合了形态学质心法,对瞳孔和虹膜在感兴区域内进行定位,降低了搜索空间,相较几何算法提升了定位准确度的同时大大缩短了算法的耗时,提高了效率。(2)提出了一种基于双红外波段的活体虹膜检测算法,通过活体人眼中血管内组织与伪造样本在不同红外波段的吸收反射率的差异来进行真伪区分。通过对活体和伪造样本中的血管纹理特征在不同红外波段下的成像清晰程度进行统计实验,选取活体人眼与伪造样本前后纹理特征数量变化差异性最大的两个红外波段作为算法中的两个对照波段。该方法可以很好的对前一算法无法很好防御的移动智能设备等新生攻击手段进行防御。针对上述方法,本文在CASIA v1.0和v2.0虹膜库及所采集虹膜库上进行了充分的实验,进一步验证了所提出方法具有一定的有效性。
[Abstract]:With the increasing trend of artificial intelligence and image recognition in recent years and the explosive growth of the number and popularity of intelligent electronic devices, biometric technology has really been integrated into people's lives. With the rapid development of iris recognition technology, experts and scholars at home and abroad have put forward and improved many classical algorithms through unremitting efforts. But no matter how advanced the recognition technology is, it can not be called qualified recognition technology if it can not effectively defend against forgery attacks. In this paper, aiming at the limitations of the existing in vivo detection algorithms based on pupil reflection characteristics, combined with iris texture change detection to form classification features to defend the corresponding attack model. According to the new attack methods such as mobile intelligent devices, which can not be well protected by this algorithm, a living iris detection algorithm based on double infrared band is proposed. It enriches and perfects the defense methods for different attack methods. The innovative work of this paper can be summarized as follows: (1) an in vivo detection algorithm based on iris texture and pupil reflection is proposed. Based on the detection of pupil light reflection characteristics and the detection of iris texture features, the original algorithm defends the model attack of simulating pupil contraction change by displacement change and so on. In order to segment the pupil more accurately, a step-by-step adaptive threshold selection algorithm is proposed in this algorithm, which has better robustness than the fixed threshold segmentation, and provides a guarantee for the subsequent positioning accuracy. An improved location algorithm based on Hough transform is proposed, which combines the morphological centroids method to locate the pupils and iris in the sensitive region, which reduces the search space. Compared with the geometric algorithm, it not only improves the positioning accuracy, but also greatly shortens the time consuming and improves the efficiency of the algorithm. (2) A living iris detection algorithm based on double infrared band is proposed. The difference of absorption reflectivity between intravascular tissue and forged samples in living human eyes in different infrared bands was used to distinguish the true and false. Through statistical experiments on the imaging clarity of vascular texture features in living and forged samples in different infrared bands, Two infrared bands with the greatest difference in the number of texture features before and after living human eyes and forged samples are selected as the two control bands in the algorithm. This method can defend against new attacks such as mobile intelligent devices, which can not be well protected by the previous algorithm. In view of the above methods, sufficient experiments have been carried out on CASIA v1.0 and v2.0 rainbow film libraries and the collected rainbow film libraries, and the effectiveness of the proposed method has been further verified.
【学位授予单位】：北方工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41
本文编号：2484050
[Abstract]:With the increasing trend of artificial intelligence and image recognition in recent years and the explosive growth of the number and popularity of intelligent electronic devices, biometric technology has really been integrated into people's lives. With the rapid development of iris recognition technology, experts and scholars at home and abroad have put forward and improved many classical algorithms through unremitting efforts. But no matter how advanced the recognition technology is, it can not be called qualified recognition technology if it can not effectively defend against forgery attacks. In this paper, aiming at the limitations of the existing in vivo detection algorithms based on pupil reflection characteristics, combined with iris texture change detection to form classification features to defend the corresponding attack model. According to the new attack methods such as mobile intelligent devices, which can not be well protected by this algorithm, a living iris detection algorithm based on double infrared band is proposed. It enriches and perfects the defense methods for different attack methods. The innovative work of this paper can be summarized as follows: (1) an in vivo detection algorithm based on iris texture and pupil reflection is proposed. Based on the detection of pupil light reflection characteristics and the detection of iris texture features, the original algorithm defends the model attack of simulating pupil contraction change by displacement change and so on. In order to segment the pupil more accurately, a step-by-step adaptive threshold selection algorithm is proposed in this algorithm, which has better robustness than the fixed threshold segmentation, and provides a guarantee for the subsequent positioning accuracy. An improved location algorithm based on Hough transform is proposed, which combines the morphological centroids method to locate the pupils and iris in the sensitive region, which reduces the search space. Compared with the geometric algorithm, it not only improves the positioning accuracy, but also greatly shortens the time consuming and improves the efficiency of the algorithm. (2) A living iris detection algorithm based on double infrared band is proposed. The difference of absorption reflectivity between intravascular tissue and forged samples in living human eyes in different infrared bands was used to distinguish the true and false. Through statistical experiments on the imaging clarity of vascular texture features in living and forged samples in different infrared bands, Two infrared bands with the greatest difference in the number of texture features before and after living human eyes and forged samples are selected as the two control bands in the algorithm. This method can defend against new attacks such as mobile intelligent devices, which can not be well protected by the previous algorithm. In view of the above methods, sufficient experiments have been carried out on CASIA v1.0 and v2.0 rainbow film libraries and the collected rainbow film libraries, and the effectiveness of the proposed method has been further verified.
【学位授予单位】：北方工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41
【参考文献】
相关期刊论文 前10条
1 铁兴华;李长栋;孙建军;荔志云;;不同光线下研究对象的年龄、性别与瞳孔变化的临床研究[J];中国民康医学;2016年14期
2 ;瞳孔大小变化之谜[J];中国眼镜科技杂志;2014年24期
3 王月明;赵士伟;张如彩;;基于人机交互的虹膜图像采集系统设计[J];中国安防;2014年17期
4 郑英娟;张有会;王志巍;张静;范胜娟;;基于八方向Sobel算子的边缘检测算法[J];计算机科学;2013年S2期
5 雷登峰;郑群辉;;一种精确的相机景深计算方法[J];信息技术;2013年08期
6 宋辉;陈浩杰;张立华;;基于灰度直方图最小跨度阈值法的瞳孔分割[J];中国印刷与包装研究;2011年02期
7 田启川;张润生;;生物特征识别综述[J];计算机应用研究;2009年12期
8 王一丁;蒋小森;;基于梯度增强的新闻字幕分割算法[J];计算机辅助设计与图形学学报;2009年08期
9 江明;刘辉;黄欢;;图像二值化技术的研究[J];软件导刊;2009年04期
10 蒋婷;谭跃刚;刘泉;;基于SOBEL算子的图像清晰度评价函数研究[J];计算机与数字工程;2008年08期
相关博士学位论文 前1条
1 何孝富;活体虹膜识别的关键技术研究[D];上海交通大学;2007年
相关硕士学位论文 前6条
1 张羝;基于多光谱的手背静脉活体检测[D];北方工业大学;2016年
2 刘博;结构相似性图像质量评价方法研究[D];大连理工大学;2012年
3 孙业超;虹膜识别预处理算法研究[D];山东大学;2011年
4 白涛;虹膜识别预处理和特征识别算法研究[D];吉林大学;2009年
5 骆丽;实时虹膜图像质量评估的算法研究与实现[D];电子科技大学;2008年
6 韩瑜;虹膜图像的质量评估方法研究[D];哈尔滨工程大学;2006年
，本文编号：2484050
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2484050.html
