第
45
卷第
2
期
2012
年
2
月
天津大学学报
Journal
ofTianjin
University
一种复杂背景下非规则带状区域的分割算法
王萍
1
,冯卫家
1
,屈展
l
,解以扬
2
,贾惠珍
2
(1.天津大学电气与自动化工程学院,天津
300072;
2.
天津市气象局,天津
300074)
Vo
l.45 No.2
Feb.2012
摘
要
z
边界层辐合线具有很强的气象风暴灾害的预示能力,它在多普勒霄达图像上表现为一条蜿蜒的、可能有岔
路的、宽度不定的、取位变化的、若隐若现的带状区域.为此,提出一种"探测窗"协同加深"印象"算法,用于
提取这种具有复杂背景的非规则带状区域.其中,探测
l
窗探身于复杂对象的局地,用以击中疑似局部目标区,并将
区域形态、取值等特点纳入探测准则之中,加深"印象"法有利于进一步地去伪存真.实验表明,该算法可以将落
入佳域的各种边界层辐合线提取出来.
关键词:图像分割;形状识别;边界层辐合线;探测窗
中图分类号:
TP39
1.
4 , TP75
1.1
文献标志码
:A
文章编号:
0493-2137(2012)02-0135-05
Segmentation Algorithm for Irregular
Be
It
-Shaped Region
Within Complex Background
WANG
Ping
1
,
FENG
Wei-jia
1
, QU
Zhan
1
, XIE
Yi-yang
2
,
JIA
Hui-zhen
2
(1. School
of
Electrical Engineering and Automation , Tianjin University, Tianjin 300072, China;
2.
Tianjin Meteorological Bureau, Tianjin 300074, China)
Abstract:
PBL
convergent lines have strong ab
i1
ity to predict weather storms, which are belt-shaped regions
with
winding, possibly branching, uncertain width and changing value in the
Doppler
radar
images. In the
paper
, an
algorithm including "detection window"
and
deepening "impression" was proposed for extracting the irregular be -
shaped region from complex background.
The
detection window is set in local areas with complex object, hit them,
and
tum
their features such as shapes
and
values into the detection criteria.
The
method
of
deepening "impression"
helps eliminate false
PBL
convergent lines
and
keep true ones.
The
experimental result shows that various
PBL
con-
vergent lines falling into the range
can
be extracted
by
the algorithm.
Keywords: image segmentation; shape recognition;
PBL
convergent line; detection window
作为图像识别的基础工作,对象的分割问题倍受
学者和研究人员关注,并致力优质高效的分割效果.
就传统的图像分割算法而言,幅度分割侧重图像的灰
度等级,边缘分割侧重对象的边界方向;差影法分割
则侧重运动着的对象等等,这些方法在应对复杂背景
下非规则对象的分割问题时仍有一定的困难.
近年来,出现了一些解决较复杂问题的针对性的
新型分割算法.
Wa
昭等
[1]
利用
MS-FCM
聚类算法,
解决像素值相似但又分布于不同区域的像素点的分
离问题,成功实现了胡子与胡须这种复杂背景下的嘴
收稿日期:
20