COL 9(4), 041003(2011) CHINESE OPTICS LETTERS April 10, 2011
Indirect building localization based on a prominent solid
landmark from a forward-looking infrared imagery
Xiaoping Wang (
²²²
)
∗
, Tianxu Zhang (
ÜÜÜ
UUU
SSS
), and Xiaoyu Yang (
{{{
)
Institute for Pattern Recognition and Artificial Intelligence,
Huazhong University of Science and Technology, Wuhan 430074, China
∗
Corresponding author: wxphero2008@yahoo.cn
Received November 11, 2010; accepted December 6, 2010; posted online March 28, 2011
A novel indirect building localization technique based on a prominent solid landmark from a forward-
looking infrared imagery is proposed to localize low, deeply buried, or carefully camouflaged buildings
in dense urban areas. First, the widely used effective methods are applied to detect and localize the
solid landmark. The building target is then precisely indirectly localized by perspective transformation
according to the imaging parameters and the space constraint relations between the bu ilding target and
the solid landmark. Experimental results demonstrate this technique can indirectly localize buildings in
dense urban areas effectively.
OCIS codes: 100.3008, 100.2000, 100.2960, 100.0100.
doi: 10.3788/COL201109.041003.
Building objects are o ften used as guides for aircraft
navigation. However, the three-dimensional (3D) build-
ing localization is a difficult problem mainly b e cause of
the complexity of the scenes. Urban environments are
very dense and composed of many typ e s of buildings,
making their analysis difficult. Thus, the reliable and
efficient recognition of buildings is crucial for enabling
such functionality. Moreover, building recognition from
forward-looking infrared (FLIR) image sequence s with a
cluttered background is a challenging task. Traditionally,
there are two modes to detect small targe ts in infrared
(IR) images. One is the direct mode, which detects
building targets directly. The other is the indirect mode,
which first selects the prominent building objects as the
solid landmark, and then detects the landmark, and fi-
nally localizes the targets. Generally, in the direct mode,
most of the recent work in building detection has focused
on the digital elevation models (DEMs), multiple aerial
images, and generic models. The methods
[1,2]
based on
DEMs are motivated by the fact that they already pro-
vide a geometric description of a scene from an aerial
imagery or airborne scanner data. Multiple view aerial
images are also the most common input data for building
extraction
[3−6]
because such data are rich in terms of 3D
information and allow the extraction of 3D primitives.
Generic model-based building reconstruction techniques
that combine several kinds of primitives are proposed
by Baillard et al.
[7]
and Taillandier et al.
[8]
, whereas the
geometric model-based building recog nition method is
proposed by Yang et al.
[9,10]
for recognizing buildings
from FLIR.
However, the DEMs, multiple aerial images, or generic
model-based building extraction metho ds are always
employed to extract visible or pr ominent buildings in
downward-looking or forward-looking scenes. If the
building is low, deeply buried, or carefully camouflaged
in dense urban areas, the direct detection methods men-
tioned above are inefficient. Moreover, some building
target enhancement techniques can be a dopted to im-
prove imaging quality and targ e t detection probability
in urban areas
[11,12]
.
In this letter, which aims to overcome the disadvan-
tages or enhance the performance of the direct mode
approaches, an indirect mode appro ach based on a solid
landmark is proposed to localize buildings in dense ur-
ban scenes. Experimental results show that the proposed
method can largely improve the direct building detection
performance of the existing widely used methods.
In dense urban scenes, if the targets are camouflaged
or occluded by their surrounding objects, the direct nav-
igation methods will not work. Therefore, indirect nav-
igation techniques are adopted; that is, solid landmarks
are selected to navigate the aircraft, a s the landmark
selection criterion is of utmost importance in aircraft
navigation. According to the extensive experiments and
detailed analy sis, the selected solid landmarks sho uld be
prominent and practical to the ensure high localization
probability of the aircraft. The criteria are as follows:
1) The solid landmark is not occluded or camouflaged
by its peripheral solid object in the flying course of the
aircraft.
2) The local contrast of the solid landmark is higher
than its peripheral 3 D solid object in the IR aerial image.
3) The real size of the solid landmark is larger or taller
than its peripheral solid object in the flying course of the
aircraft.
To localize the building target precisely in the IR im-
age, the space constraint relations between the building
target and the solid landmark in the actual scene, which
Fig. 1. Sketch map of the distance between the target and
the solid landmark in DOM.
1671-7694/2011/041003(4) 041003-1
c
2011 Chinese Optics Letters
资源评论
weixin_38737565
- 粉丝: 7
最新资源
资源上传下载、课程学习等过程中有任何疑问或建议，欢迎提出宝贵意见哦~我们会及时处理！
点击此处反馈
