Many thanks to Giuseppe Maldera, of our partner company Planetek Italia, for providing the following Spatial Model
Coastlines, shoals and reefs are some of the most dynamic and constantly changing regions of the globe. Monitoring and measuring these changes is critical to marine navigation and an important tool in understanding our environment.
|Extracted coastline over a DigitalGlobe WorldView-2 image. Polyline result on the left (in red) and polygon on right (blue)|
|The screenshot above includes copyrighted material of DigitalGlobe, Inc., All Rights Reserved.|
Following the scientific approach of this paper (Maglione, 2014) and using the ERDAS IMAGINE 2018 Spatial Modeler, Planetek have built a geoprocessing model which automatically extracts the instantaneous coastline, as identified in WorldView-3 (or -2) imagery, without considering tide (i.e. the land/water interface as depicted in the satellite image at the time of acquisition).
To enhance geometric resolution of the results, pan-sharpening should preferably be applied so as to obtain maps with the same pixel dimensions and accuracy of the panchromatic data.
The shoreline identification is based on grouping all the pixels from the imagery into three contrasting classes: sea, land and vegetation. The groups are produced using the Normalized Difference Water Index (NDWI) and statistical clustering.
Thanks to the great functionalities of the Spatial Modeler every step is connected and completely automatic. After the classification the raster is converted into features (vector data) and then the features are cleaned up and optimized using new "geometry manipulation" operators, available in ERDAS IMAGINE 2018. The user may optionally constrain the processing extent using an AOI or bounding box, and then run the model, resulting in two shapefiles, one polygonal and one linear.
|Extract Coastline sub-model|
|Merge and Measure Area sub-sub-model|
|Sea Selection sub-sub-model|
|Attach Attributes sub-sub-model|
The image must contain bodies of water. If little or no water is present, the results will be erroneous.
The model also assumes that the sea area is the largest area of water in the image. If your image instead contains just a short section of coastline and a large inland lake, for example, it will extract the lake. This is a good scenario where using the Boundary or AOI inputs would help capture the desired feature.
The model is generally designed for WorldView-2 and -3 imagery. Preferably pan-sharpened, but it's not absolutely necessary. However the input image must have the Coastal Blue wavelength in Band 1 and the NIR2 wavelength in Band 8. Other images (such as Sentinel-2) that capture in similar wavelengths could theoretically be processed using the same technique, but the model will need to be modified to point to the bands appropriate to those sensors.
Pasquale Maglione, Claudio Parente & Andrea Vallario (2014) Coastline extraction using high resolution WorldView-2 satellite imagery, European Journal of Remote Sensing, 47:1, 685-699, DOI: 10.5721/EuJRS20144739
Link to this article: https://doi.org/10.5721/EuJRS20144739
Worldview Pan-Sharpened: Provide an input filename for the image to be processed. The image should preferably be pansharpened to provide the highest level of spatial fidelity. However it can be a simple MS image if desired. Band 8 must be the NIR2 wavelength data and Band 1 the Coastal Blue.
Boundary Input (Optional): Optionally the user can provide a map extent (e.g. by using the Inquire Box dialog) to constrain the processing. If the entire image is to be processed this field can be left blank.
AOI Input (Optional): Optionally the user can provide an Area of Interest (e.g. by digitzing and selecting an AOI polygon in a 2D View) to constrain the processing. If the entire image is to be processed this field can be left blank.
Minimum Island Size (Optional): Specify an area (in the units of the WorldView Pan-Sharpened input image). Any holes in the sea area polygon (such as those caused by islands, sun-glint, etc) that are less than this area will be removed. By default the field is blank and all holes will be removed. If your coastline is complex with large near-shore islands (e..g in salt marshes, artificial harbors, river deltas, etc) you may wish to specify a minimum area threshold above which these islands are retained.
Classified NDWI OUT (Required): This output is required. A raster filename must be provided to write the three-class thematic raster derived from the Normalized Difference Water Index.
NDWI OUT: If a filename is provided, the Normalized Difference Water Index values will be written to a floating-point raster file
Linestring Coastline OUT: If a filename is provided, the extracted coastline will be produced as a line geometry polyline.
Polygon Coastline OUT: If a filename is provided, the extracted coastline will be produced as an area geometry polygon.