Sunteți pe pagina 1din 2

Why is separating the shadow and water information hard in multispectral image

The presence of shadows in optical satellite images limits the application of remote-sensing technology. Optical
remote sensing is critically important for a variety of applications such as land-use and land-cover change detection,
natural resource assessment, and urban structure quantification. However, the quality of optical image data is often
degraded by shadows that are inevitably formed by complex surface terrains and atmospheric conditions. Shadows
in remote-sensing images misrepresent image information, which can lead to potential errors in the derivation of
surface parameters such as surface reflectance and reflectance-based indices. In addition, the presence of shadows
can lead to misclassification and misinterpretation when performing information extraction and change detection
with remote-sensing images. Therefore, shadows should be corrected prior to classification and change detection of
remotely sensed imagery. Shadow correction processes normally consist of two sequential steps: shadow detection
(or extraction) and shadow restoration (or removal). Tremendous progress has been made in shadow detection and
restoration.

Urban surface water mapping is essential for studying its role in urban ecosystems and local microclimates.
However, the fast and accurate extraction of urban water remains a great challenge due to the limitations of
conventional water indexes and the presence of shadows. Urban surface water such as rivers, lakes, reservoirs, and
ponds, exerts a significant influence on urban ecosystem services and local microclimates. As a consequence of
Land Use/Land Cover (LULC) and environmental changes and natural hazards, variations in urban surface water
may result in a series of ecological, climate, health, and socioeconomic problems, such as water supply shortages,
biodiversity losses, aggravation of the urban heat island effect, and even outbreaks of waterborne infectious
diseases. These problems tend to be more prominent in cities with rapid urbanization. With the development of
urbanization, the urban space has been expanded, leading to the shrinking of water bodies. Meanwhile, frequent
human activities may lead to the deterioration of water quality towards being turbid, stink, or black. Therefore,
timely and accurate mapping of urban surface water is crucial for urban planning and disaster assessments. Remote
sensing techniques, with their advantages of large area coverage, integration, speed, and periodicity, have been
widely used to delineate surface water and monitor surface water dynamics. Various methods have been proposed to
identify surface water bodies, which can be divided into four types: thematic classification, spectral unmixing,
single-band thresholding, and spectral water index. The last is the most widely used, due to its ease of use, relatively
high mapping accuracy, and low computational expense. Over the past few decades, many water indexes have been
presented in the literature. McFeeters proposed the first water index called the Normalized Difference Water Index
(NDWI) with a default threshold of 0, which utilized the reflectance difference between water and vegetation and
soil in the red and near-infrared (NIR) bands. To suppress the signals from buildings, Xu replaced the NIR band
with the shortwave infrared (SWIR) band in the NDWI formulation, creating the Modified Normalized Difference
Water Index (MNDWI). Although the MNDWI shows high accuracy, it was still unable to suppress shadows.
Therefore, Feyisa et al. proposed an automated water extraction index (AWEI) and that has been demonstrated to be
effective in different environments, particularly in mountainous areas with deep shadows.
Previous work on water extraction mainly faced two difficulties.

1. It is difficult to obtain an accurate position of water boundary because of using low-resolution images.
2. Like all other image-based object classification problems, the phenomena of “different objects same
image” or “different images same object” affects the water extraction.

Shadow of elevated objects (e.g. buildings, bridges, towers and trees) scattered in the remote sensing image is a
typical noise object for water extraction. In many cases, it is difficult to discriminate between water and shadow in a
remote sensing image, especially in the urban region. Recently, a number of new approaches to water extraction for
remote sensing images of different sources have been proposed. The object-oriented technique is a new technology
which appeared recent years, rather than in pixel-level, it classifies remote sensing images in patch-level.
Afterwards, various features such as textural features, shape, spatial relations and reflectance statistics were applied
into water extraction. Water extraction is one of the important measures to protect and monitor water resources on
earth. We feel that there is a clear need for more efforts in water extraction for effectively maintaining the water
resources. The process of water extraction includes two hierarchies:

(1) image segmentation process and water extraction

(2) shadow removal process.

With the progress of remote sensing, more and more high-resolution images are used for city planning and
administration. In these images, shadows are cast by buildings present, especially in urban environments. Shadows
can be used to detect buildings and estimate the heights of the buildings. On the other hand, they involve loss of
information for the surface under the shadow’s present difficulties for image interpretation, image matching, change
detection and other applications. Shadow removal is a critical problem in image processing. In image analysis, field
researchers employ retinex theory to restore the shadows in their work. The problems are parameters for the shadow
edges and computational efficiency. For an image of 4806640 pixels, it takes about 10 minutes for a 333M Pentum
computer to remove a shadow area approximately a quarter of the size of the image. In remote sensing only a few
works on shadow detection have been carried out. Recently, some works on shadow detection at a very high spatial
resolution like Ikonos have been published. Some methods used principally colour and spectral properties to detect
shadow. In some papers, binocular image computations have been used to remove the shadows. A mathematical
model was proposed by Zhou to detect the occlusion by visibility analysis and photogrammetric engineering. When
the building model is given it is projected onto the image and the coordinates of the corners of the image are
calculated. Shadow area is fulfilled by the same position of the slave image acquired meanwhile from the master
image.

S-ar putea să vă placă și