Select to view content in your preferred language

Water/wet area detection in 8-band WorldView 2 imagery

444
1
08-15-2013 03:47 PM
Will
by
Frequent Contributor
Hi,

I'm trying to identify standing water/wet areas on a WorldView 2 image with the following bands:
50 cm:  Panchromatic
2 meters:  Coastal, Blue, Green, Yellow, Red, Red Edge, NIR1, NIR2

The location is a mountainous area with pine trees and meadows.  I'm also trying to differentiate water from shadows.  Any suggestions on approaches?

Thanks much
0 Kudos
1 Reply
larryzhang1
Occasional Contributor
Hi,

I'm trying to identify standing water/wet areas on a WorldView 2 ...also trying to differentiate water from shadows.  Any suggestions on approaches?

Thanks much


William,

If understanding correctly, you want to 'extract' information (lakes, rivers) from image?  If so, the solutions vary, because each solution relies on the output that you are expecting.

For your case, even though three kinds of feature are similar by NIR(s) (including water, shadow, and vegetation), spectra values are still different with NVDI index.

So, if satisfied with a �??raster�?? mapping on water body, visible band against the near-infra band (NDVI) may be sufficient to help you to isolate those (lakes, river) and others.

With ArcGIS, you can create MD first (that is, to pansharpen the WV-2 MS with PAN by default via Raster Type: World View -2), and then apply this MD with NVDI within �??Image Analysis�??. Generally, you will easily and quickly get color-coded indexing map from 0 ~255, so that you can conveniently export desired values (water mostly between 10~70, for your case, maybe the range change) for further editing�?�

Of course, if planning to more accurately extract features (lakes, rivers, meadows, forest stands), saying, to update GIS database, it is very often to use more advanced algorithms like object-oriented feature extractions, mainly from other solution vendors like ERDAS �??Objective�?? or ENVI �??Feature Extraction�??;


++++++++++
By the way, if you just want to 'enhance' image to differentiate objects (water, shadow) in image, it would be easy for you just to play 'function chain' within MD, including histogram stretching, and/or NIR ratio operating ...
0 Kudos