IEEE Geoscience and Remote Sensing Magazine - September 2019 - 66

SPATIAL CHALLENGES

However, there are relatively few reviews of UIS detection
in this literature. A survey of urban heat island assessment
and surface energy flux estimation by thermal infrared (TIR)
remote sensing techniques examined methods, applications,
and trends in applying TIR data for urban heat island and
energy balance modeling studies [17]. A systematic review of
UIS studies examined the data requirements for UIS, assessed
various methods of UIS extraction and estimation, and proposed future developments in UIS estimation and mapping.
The overview highlighted a classification of UIS methods,
including those that are pixel based, subpixel based, object
based, and artificial neural network (ANN) based [18]. In assessing the impact of urbanization on global environmental
change via remote sensing mapping, the work in [19] described data source selection methods and the limitations of
mapping related to UIS. The study also focused on mapping,
urban indices, and modeling of urban remote sensing.
A survey of the methods of urban extent mapping used
time-series nighttime light data, such as from the Defense
Meteorological Satellite Program/Operational Linescan
System (DMSP/OLS). This review systematically examined
three key aspects in the intercalibration framework of nighttime light data [20]. However, few reviews have considered
the challenges of UIS detection, especially systematic problems. Moreover, as more RSIs and new algorithms are used
for UIS detection, an examination of the established algorithms is required.

SPATIAL RESOLUTION
Spatial resolution is based on the minimum unit of recognizable ground objects in RSIs. The actual size of the spatial
resolution value on the ground is called the ground resolution. If this resolution is lower than the spatial resolution,
the ground objects are difficult to detect in images unless
they have a distinctive brightness (color) or a unique edge
and textural features [21], [22]. Generally, remotely sensed
data can be classified into three types according to their
spatial resolution: coarse, medium, and high [20]. Coarse
spatial-resolution data (>100 m), such as the MODIS data
sets, are applied in large-scale imaging with higher temporal resolution [23]-[27]. However, coarse spatial-resolution
data have a lower UIS detection accuracy for local and regional mapping [28], [29] and are used as auxiliary data because of their high temporal resolution [30].
Medium spatial-resolution data (>10 m and <100 m) are
used for regional UIS detection [31], such as data derived
from SPOT 4 and Landsat [32]. These data sets have a large
number of mixed pixels because of complex ground coverings that influence the accuracy of ground object classification [33], [34]. High-spatial-resolution data (<10  m) (e.g.,
from WorldView, IKONOS, and QuickBird) are not suitable
for global-scale UIS detection with current technology because of costs associated with mass data acquisition, storage, and processing complexity [35], [36]. Moreover, highspatial-resolution data contain many shadows, caused by
shade from high buildings or trees, influencing the accuracy of UIS detection [37], [38].
Figure 2 shows the effects of RSIs with different spatial
resolutions at the same geographical location (central longitude and latitude: 32.19°  N, 118.26°  E). The higher the
spatial resolution, the easier it is to identify ground object
types. Images with high spatial resolution have clear feature
boundaries, an aspect that is more helpful in UIS extraction.

CHALLENGES IN URBAN IMPERVIOUS
SURFACE DETECTION
Difficulties in UIS detection have arisen from various application requirements. In terms of the extent and frequency
of their influence, these challenges can be divided into four
categories: spatial (e.g., spatial resolution and urban boundary confusion), passive optical remote sensing (e.g., spectral
confusion and issues relating to seasonal changes), active remote sensing, and multisource remote sensing techniques.

(a)

(b)

(c)

FIGURE 2. A comparison of RSIs with different spatial resolutions at the same geographical location. (a) MODIS (250 m), 14 March 2016; (b)

Landsat OLI (30 m), 6 June 2018; and (c) IKONOS (4 m), 18 June 2009.
66

IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

SEPTEMBER 2019
IEEE Geoscience and Remote Sensing Magazine - September 2019

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