An Effective Morphological Index in
Automatic Recognition of Built-up Area
Suitable for High Spatial Resolution
Images as ALOS and SPOT Data
Bo Yu, Li Wang, Zheng Niu, and Muhammad Shakir
Abstract
Building detection from remote sensed images is the main
technique to monitor economic or environmental develop-
ment of an area. Advanced Land Observing Satellite (
ALOS
)
and
SPOT
data are reliable sources due to the limitation of
weather, position, time, and other practical reasons. How-
ever, to the best of our knowledge, algorithms proposed in
the identification of buildings mostly aim only at images with
very high spatial resolution or high spectral resolution. There
are few algorithms for detecting buildings from
ALOS
and
SPOT
data. A built-up detection index (
BDI
) is proposed in this
paper to automatically identify buildings from images with
10 meters resolution. It synthesizes morphological theory and
normalized differential vegetation index (
NDVI
) to enhance
buildings by suppressing vegetation. Four images of
ALOS
and
SPOT
are used to verify the efficiency, stability and ac-
curacy of
BDI
. Experiments show that
BDI
is suitable to detect
buildings from 10 meters resolution with reliable accuracy.
Introduction
Urbanization is one of the most significant indicators of
economy development. Furthermore, buildings construction
is one of the most obvious indicators showing development
of an area. Therefore, monitoring distribution and the amount
of buildings in a specific period can be helpful in evaluating
the rate of development of an area (Champion
et al.,
2010).
Apart from that, urban mapping and urban extension (Grey
et al
., 2003; Nemmour and Chibani, 2006) are two other main
application areas for building-detection as well.
Multiple sensors, such as Thematic Mapper (
TM
) (Li
et al
.,
2011; Tian
et al
., 2012), Advanced Land Observing Satellite
(
ALOS
) (Liu
et al
., 2011; Zhao
et al
., 2012) and
SPOT
(Yang
et
al
., 2011) have expanded the applications of remote sensing
in object recognition. Compared with very high spatial resolu-
tion images, such as unmanned vehicle images and Geo-Eye,
ALOS
, and
SPOT
images have relatively low spatial resolution
but more bands, which enhanced its ability to distinguish
buildings. For
ALOS
and
SPOT
images, buildings are difficult
to be discriminated from other built-up objects, such as roads
and open spaces. If buildings are of small size, it would be
difficult to distinguish them from other features and mostly
take the form of two or three pixels. Moreover, the area cov-
ered by
ALOS
and
SPOT
image is largely enlarged and more
types of surface features are included. Therefore, an algorithm
with high robustness and less consuming time is demand-
ing. However,
ALOS
and
SPOT
data include near infrared (
NIR
)
band. Due to
NIR
, it is possible to detect vegetation more accu-
rately, which is because vegetation has high reflectivity in the
spectral range of
NIR
. The methods for detecting built-up area
from
ALOS
and
SPOT
data are still not well explored.
Masek
et al
. (2000) adopted normalized differential vegeta-
tion index (
NDVI
)-differencing approach to separate vegeta-
tion from built-up area with an accuracy of 85 percent. Logic
tree was used by Xu (2002) and Xian and Crane (2005) to
recognize built-up territory with almost the same accuracy of
85 percent. However, both algorithms need much time and
memory to run, which would largely decrease the efficiency
with an image of large size. As for simple algorithms based on
indices only, a few have been proposed. Xu (2008) has come
up with (index-based, built-up index)
IBI
for effectively tell-
ing built-up area from satellite images with impressive rapid
extraction. Unfortunately, the index needs Mid-Infrared (
MIR
)
band, which is sometimes unavailable. Moreover, a manually-
decided threshold is required to extract buildings, which
lowers robustness of
IBI
.
Huang and Zhang (2011) have proposed an index called
Morphological Building Index (
MBI
) based on the reconstruc-
tion morphological operation presented by Pesaresi and
Benediktsson (2001) for automatically detecting buildings.
However, it needs much postprocessing after the recogni-
tion. High omission and commission errors are obstacles in
application. Differential morphological profile (
DMP
) and its
derivative methods (Benediktsson
et al
., 2003; Pesaresi and
Benediktsson, 2001) have been proposed to detect buildings
from high spatial resolution images. They are well-known for
high automation and easy access, but large consuming time is
a key drawback in practical application.
In this study, a novel morphological index is proposed
to detect buildings. It is based on a combination of morpho-
logical theory and normalized difference vegetation index.
Bo Yu and Muhammad Shakir are with the the State Key
Laboratory of Remote Sensing Science, Institute of Remote
Sensing and Digital Earth, Chinese Academy of Sciences,
Beijing 100101, China; and the Graduate University of
Chinese Academy of Sciences, Beijing 100049, China.
Zheng Niu is with the 2Graduate University of Chinese Acad-
emy of Sciences, Beijing 100049, China (
.
Photogrammetric Engineering & Remote Sensing
Vol. 80, No. 5, June 2014, pp. 529–536.
0099-1112/14/8006–529
© 2014 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.80.6.529
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
June 2014
529
