Abstract:
Flood monitoring has enjoyed a recent wave of increased interest fueled by the need
of disaster management, risk/damage assessment and rehabilitation process activities.
Flood disaster management activities are comprised of flood mapping and flood forecasting.
Unsupervised flood monitoring techniques have many distinct advantages over
other modalities such as supervised and semi-supervised techniques. Unsupervised
flood mapping (which includes pre and post disaster dataset comparison) is usually
preferred because it is cost effective and fast.
Unsupervised flood monitoring has been extensively studied over the past decades
in various dimensions (the development of geographic information systems, integration
of digital elevation models, change detection techniques using fuzzy logic, the use of
Satellite dataset and ariel imagery etc) resulting in a dramatic improvement. However,
flood monitoring severely degrades due to presence of cloud and lightening in dataset,
availability of dataset and associated labour cost. Radar remote sensing provides great
contributions in flood monitoring, however, they suffer from the issue of poor contrast
and speckle noise.
A contrast enhancement based flood monitoring technique is proposed for the visualization
of Synthetic Aperture Radar imagery. The proposed technique is composed
of three steps (adaptive histogram clipping, re-mapping and contrast enhancement)
based flood monitoring. The contrast enhancement is further composed of other techniques
(histogram stretching, histogram smoothness and adjustable histogram equalization).
The techniques overcomes the issues of over enhancement, unusual artifacts
and over detection of flooded areas.
The proposed techniques generated useful results, however, undated and inundated
areas are not clearly differentiable, which are further improved by the use of proposed
change detection approach. A quality difference image is generated using proposed
image fusion of mean and log ratio images (generated using pre and post flooded images).
The technique overcomes the issue of misclassification of inundated areas and
minimize the use of complex change detection algorithms. Proposed techniques will
bring improvement in state-of-the-art techniques.
