第
42
卷第
4
期
2006
年
7
月
南京大学学报(自然科学)
JOURNAL
OF
NANJING
UNIVERSITY
CNA
TURAL
SCIENCES)
自动指纹识别系统预处理技术
及细节特征提取算法的研究祷
杨小冬，宁新宝气尹义龙
(南京大学电子科学与工程系，生物医学电子工程研究所，南京，
210093)
Vo
l.
42 ,
No.4
J
uly
,
2006
摘
要:
指纹图像预处理是自动指纹识别系统
CAFIS)
的关键步骤，它的好坏直接影响到整个系统的
速度和准确率.详细介绍了预处理过程中指纹图像与背景分离、方向信息提取、纹线提取、二值化、细
化，以及细节特征提取等关键技术.在这些模块中，提出了一些改进算法和新的思想.根据人指纹本身是
一个曲面结构的特点，采用一种自适应的局部阂值分离方法，解决处于不同深浅区域的指纹图像均能有
效地同背景分离;提出了一种改进的、基于非彻底细化图像的细节特征提取算法，在不对纹线进行任何
处理的情况下，直接从细化指纹图像上提取原始细节特征点集合，并针对各种噪声产生的伪特征点进行
修复;另外，还介绍了一种基于灰度纹线跟踪直接提取指纹图像细节特征的新方法，并与传统算法进行
了比较.这种方法不经过二值化、细化等过程，能显著减少预处理带来的累计误差，加快处理速度，在一
定程度上提高了准确率.实验证明，以上算法行之有效，在实际应用和严格测验中取得了较好的效果.
关键词:
指纹预处理，细节特征提取，纹线跟踪
中图分类号
TP
391
Fingerprint
Image Preprocessing Technique
and
Feature Extraction Algorithm
Yσ
ng
Xiao-Dong
,
Ning
Xin-B
α
0
，
Yin
Yi-Long
CDepartment of Electronic Science and Engineering,
Institute
of Biomedical Electronic Engineering,
Nanjing
University
, Nanjing, 210093 ,
China)
Abstract:
The
biologic identification technology
that
is delegated by
fingerprint
has
widely developed in recent
years.
It
is recognized to
bring
a revolution in identification area.
The
study
of
fingerprint
has
high academic vague
as well
as
social benefit,
and
is becoming
one
of
the
research
hotspots
in
many
countries.
Of
the
whole
Automatic
Fingerprint
Identification
System
C
AFIS)
,
fingerprint
image preprocessing is
the
core
and
key
step.
The
input
fingerprint image
should
be
preprocessed
first
before feature
extraction
and
matching
can be done afterwards.
These
processes
can
influence
the
speed
and
accuracy of
the
entire
syste
m.
In
this
paper
,
we
made
a deep
research
on
these
preprocessing
techniques,
which
included
the
separation
of
foreground
and
background
regions,
the
orientation
information
extraction
, ridge
extraction
, image enhancement,
binarization, image
thinning
, as well as feature extraction.
In
these
process
modules
,
we
brought
forward
some
new
algorithms
and
novel
thoughts.
Therefore
,
the
AFIS
was
complemented and perfected in several aspects. According
to
the
characteristic
structure
that
the
fingerprint
itself is a curving plane,
we
introduced
a
separating
method
using
self-adaptive local
threshold
, which could solve
the
problems
that
fingerprint images in
areas
of
different
depth
could
养基金项目:南京大学重大应用研究预研项目
(2001-03)
收稿日期
:2005-12-15
柑通讯联系人.
E-mail: xbning@nju.edu.cn
剩余10页未读，继续阅读
资源评论
