第26耋芝12期
计
算 机 学
报
v0I.26
N。.12
1111兰!!旦
塞坚!些!!兰!竺坚曼些垒兰呈!呈望坚!坚三兰里!
呈!!:!!!!
一种基于球面透视投影约束的鱼眼镜头校正方法
英向华
胡占义
(中国科学院自动化研究所模式识别国家重点实验室北京10。。80)
摘要
鱼眼镜头摄像机具有较大视场,但是,使用鱼眼摄像机拍摄的图像会有非常严重的变形该文研究基于球
面透视投影约束的鱼眼镜头校正方法.球面透视投影约束是指空间直线的球面透视投影为球面上的大圆.作者首
先使用含有变形校正参数的鱼眼变形校正模型,将空间直线的龟眼投影曲线上的点映射为球面点,然后通过球面
点剑大圆的球面距离最小来拟合大圆,恢复了变形校正参数,从而实现了鱼跟图像的校正.模拟实验和真实图像实
验表明.该方法能得到比较满意的校正结果.
关键词
鱼服镜头;基于图像的绘制;变形校LE;球面几何
中图法分类号Tp391
Fisheye
Lense
Distortion
Correcti佃
Using
Spherical
Perspective
Projection
Constraint
YINGⅪa“g—Hua
HU
Zhan—Yi
(S抽“K掣L口60托f07y
o,Pd££…RP∞gHtz
z。”·J”sfl}“£P
o,AHf…£z。",(巩z”PsP
Afnd柳叫。,sf
zP眦e5,Be“z”g
100080
Abstract
Fisheye
lenses
are
often
used
to
enla。ge
the
field of
view
of
a
camera.
But
the
images
taken
wIth
the
flsheyc
camera
often
have
severe
distortions.
Thi8
pape’proposes
a
new
fisheye
lenses(1istortlon
correction
method
using
sphe“cal
perspective
projection(SPP)constraint.SDace
llne
are
genera儿y
projected
into
curves
on
a
fi3heye
image+The
SPP∞nstraint
means
that
a
sDace
Ilne
must
be
projected
into
a
great
circIe
on
a
unit
sphere
llnder
the
SPP
model.
The
reason
of
u—
si“g
the
SPP
model
ls
to
handle
the
situati。n
of
larger-than一180一degree
view
angles,and
our
case
happens
to
be
true.
(Jur
proposed
djstortion
correction
process
consists
of
the
following
two
main
steps:Firstly,polyn。miaI
models
for radial
and
ta“gential
distortions
with
free
parameters
are
selected;Secondly,thes。parameters
are
estimated
based。n
the
fact that
the
projected
image
curves
oi
space
lines
should
be
mapped
to
great
circles
on
a
un“spherical
surface,i.e.,the
SPP
constraint.By
minimizing
the
sum
of
errors,more
specifically,the
squares
of
spherical
distances
from
mapped
points
to
their
corresponding
best—fit
great
circles,the
parameLers
are
obtained.
Our
minimization
process
also
consists
of
two
steps,at
first,an
inmal
estimation
is
obtained
bv
the
1访earizat】。n
oI
the
tw。di5t。rtl。n
modeJs,th朗Levenbe‘g~Marquardt
algorithm
js
employed
f。r
the
final
non-linear
optimization.Finally,expenmental
results
with
synthetic
data
under
different
noise
levels
as
weU
as
with
real
fisheye
images
are
reported,and
the
results
appear
satisfactory.
Keywords
fisheye
knse8;imag}based
rende“ng;distorti。Ⅱcorrectio“;sphe“cal
geometry
引
言
在机器人导航、虚拟现实,基于图像的绘制以及
视觉监控等许多计算机视觉领域,需要使用具有较
大视场(field
of
view)的广角或鱼眼摄像机.但使用
广角或鱼眼摄像机拍摄的图像具有非常严重的变
形.如果我们想利用这些具有严重变形图像的透视
收稿日期:2002
10
30;修改稿收到日期:2003一06—10.奉课题得到国家自然科学基金(60175009,600330l
o)资助.英向华,男,19
76年生,
搏十研究生,主要研究方向为基于图像的绘制、计算机视觉、计算机图形学等胡占义,男,1961年生,博士,研究员,博上生导师,主要研
究方向为摄像机标定与三维重建、主动视觉、几何基元提取、视觉机器人导航、基于图像的建模和绘制.Bmail:huzy@nlpr
Ia
ac cn
万方数据