IEEE Geoscience and Remote Sensing Magazine - September 2019 - 67

An inherent relationship exists between scale and spatial resolution. Spatial phenomena on one scale may not exist on
another because of the spatial resolution. Therefore, image selection with an optimal scale and spatial resolution remains
challenging in certain research and application fields.
However, the identification and classification of UIS
from RSIs is dependent on spatial resolution and is related
to the shape, size, and relative differences of a ground object
compared with the brightness and structure of surrounding
objects [39]. High spatial resolution can decrease boundary
mixed pixels and improve the accuracy of ground object classification because of the clear shape and structure [40]. Additionally, UIS detection accuracy can be improved by making full use of RSIs' spatial (position, shape, geometry, and
boundary), spectral, temporal, and context features [41].
URBAN BOUNDARY CONFUSION
Clear urban boundaries can delineate the scale and trend
of urban development [18]. The rural-urban intersection
zone has complicated ground cover, with different materials (e.g., plastic films and color steel tiles) that possess
spectral reflectance properties similar to high-reflection
UIS. These results make it difficult to extract city boundaries, leading to an uncertain overall accuracy regarding
the UIS area. In Figure 3, a comparison of urban boundaries and urban centers with different spatial resolutions is
shown. Figure 3(a) and (b) presents medium spatial-resolution images (Landsat 8 Operational Land Imager [OLI],
30 m), while Figure 3(c) and (d) shows high-spatial-resolution RSIs (IKONOS, 4 m) appear.
Figure 3(a) and (c) shows the junction of urban and
rural areas containing extensive vegetation, water bodies,
bare soil, and low buildings, leading to unclear ground object boundaries. Figure 3(b) and (d) depicts an urban center
with dense buildings that exhibit a clear texture. Moreover,
land cover subtypes possess varying degrees of spectral reflectance, leading to ground object confusion [42]. Such
confusion is likely caused by homologous spectral reflectance of diverse ground objects in coarse to medium spatial-resolution images with mixed pixels. The decision tree

(a)

(b)

(DT) algorithm based on urban sparsity provides a solution
for urban boundary determination [43], [42].
PASSIVE OPTICAL REMOTE SENSING CHALLENGES
SPECTRAL CONFUSION
In the real world, ground objects are diverse and ever
changing, and different types of ground objects are interlaced and exhibit near homogeneity of spectral features.
Moreover, identical types of ground objects often appear
heterogeneous in different scenes [44]. According to differences in the brightness of impervious ground objects
in images, such surfaces can be divided into high-albedo
impervious surface (HIS) (e.g., new concrete roads and
parking lots) and low-albedo impervious surface (LIS) (e.g.,
asphalt roads, old concrete buildings, and galvanized iron
roofs) [18], [19]. Sands and sandy soil have high spectral
reflectance and can be confused with HIS, such as in highdensity residential regions and compact commercial areas.
Bare soil can be confused with LIS, such as high-density
low or old buildings, which are difficult to distinguish from
shrubbery [45]-[47].
Figure 4 compares the reflection values of bare soil,
sands, and UIS in a Landsat 8 OLI image. Bare soil, sands,
and UIS were selected from the images and are represented
by blue, magenta, and red, respectively. Figure 4(a) and (b)
shows true color images of Landsat 8 OLI, with Figure 4(a)
presenting bare soil and LIS, and Figure 4(b) depicting sand
and HIS. Figure 4(c) displays the reflection values of bare
soil, sand, and UIS in the rectangular frame; their reflection
values are relatively close, which can lead to confusion.
Water and shadows tend to have lower spectral reflectance
and can be easily confused with LIS [48]. To distinguish LIS
from water and shadows, TIR bands have been used, which
have clear temperature variations in LIS compared with no
significant changes in the temperature of water or shadows
[49]. The influence of bare soil on UIS detection can be altered by the change of seasons. For example, in summer,
most of the natural bare soil is covered by vegetation, resulting in a reflectivity different from that of UIS [50].

(c)

(d)

FIGURE 3. A comparison between urban boundaries and an urban center, with different spatial resolutions: (a) urban fringe area, Landsat 8

OLI (30 m); (b) downtown area, Landsat 8 OLI (30 m); (c) urban fringe area, IKONOS (4 m); and (d) downtown area, IKONOS (4 m).
SEPTEMBER 2019

IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

67
IEEE Geoscience and Remote Sensing Magazine - September 2019

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - September 2019
