Automatic Channel Network Extraction From Remotely Sensed Images by Singularity Analysis

Abstract

Abstract-Quantitative analysis of channel networks plays an important role in river studies. To provide a quantitative representation of channel networks, we propose a new method that extracts channels from remotely sensed images and estimates their widths. Our fully automated method is based on a recently proposed Multiscale Singularity Index that responds strongly to curvilinear structures but weakly to edges. The algorithm produces a channel map, using a single image where water and non-water pixels have contrast, such as a Landsat nearinfrared band image or a water index defined on multiple bands. The proposed method provides a robust alternative to the procedures that are used in remote sensing of fluvial geomorphology and makes classification and analysis of channel networks easier. The source code of the algorithm is available at:

The Delta Connectome: A network-based framework for studying connectivity in river deltas

“…The inherent complexity of distinguishing land and water in areas subject to frequent inundation and the lack of steep gradients to guide the extraction based on classic steepest descent approaches make the extraction of delta networks challenging. Recent developments in image processing offer robust solutions to this problem; the multi-scale singularity approach of Isikdogan et al (2015) has been successful in extracting channel networks in coastal systems. While more work is needed in ensuring the connectivity of the extracted network, the procedure is robust and has already been applied to other systems including numerical modeling results (Liang et al, 2016).…”

“…Over smaller spatial extents, techniques such as lidar have increased the capability of remotely capturing geomorphic processes at fine scales (Passalacqua et al, 2015). Advanced tools for the extraction of geomorphic features from remotely sensed data of coastal environments are now available (Shaw et al, 2008;Geleynse et al, 2012;Isikdogan et al, 2015) and physical and numerical modeling of deltaic systems have been underway for more than a decade. Techniques for field data acquisition have also been improved and new techniques have been developed (e.g., multi-beam sonar, green lidar, ADCP sensors), providing detailed bathymetry and flux data.…”

Quantifying the patterns and dynamics of river deltas under conditions of steady forcing and relative sea level rise

“…Note that here “land” refers to the whole delta surface, which includes both islands and channels. Channel map: a binary channel map is obtained by selecting pixels in the land map with a threshold flow velocity of 0.3 m/s, a value chosen in accordance with the minimum velocity required to transport sediment in our DeltaRCM runs (Figure c). Wetted map: a binary map of wetted surface area is created by selecting pixels in the land map that are channel pixels or have depth greater than 0.5 m (Figure d). This wet map includes active and abandoned channels, lakes, and shallow bays inside the delta. Channel centerline map: channel centerline pixels are extracted from the channel map using the method developed by Isikdogan et al [] (Figure e). Note that a binary map is not required to implement this method, as a gray scale flow map (e.g., using normalized discharge values) can also be used as input. Island map: a 3 by 3 median filter (window size selected to preserve 1‐ to 2‐cell wide channel features) is applied to the wet map to remove noise, and then islands are identified as connected pixels, which are then marked as separate groups (Figure f).…”

“…The model (DeltaRCM) is available for download at http://csdms.colorado.edu/wiki/Model_download_portal. The channel centerline extraction tool of Isikdogan et al [] is available for download at http://live.ece.utexas.edu/research/cne/. The Matlab codes for computing the metrics in Table can be downloaded from http://sites.google.com/site/passalacquagroup/home.…”

Characterization of Deltaic Channel Morphodynamics From Imagery Time Series Using the Channelized Response Variance

“…The multiscale singularity index has been demonstrated to be a highly effective way to automate the extraction of channel features from images (Muralidhar et al, ; Isikdogan et al, ) and is a combination of low‐order multiscale derivatives that are calculated on the intensity of the grayscale input image as where f 0, θ , σ , f 1, θ , σ , and f 2, θ , σ are the responses to the zero‐, first‐, and second‐order derivatives of Gaussian kernels at a given scale σ along direction θ (Figures S1a–S1c). The singularity index is designed to respond strongly to curvilinear structures, where the second derivative in the numerator is large, and to respond weakly to edges, where the first derivative in the denominator is large.…”