0

I have NIR, green, red edge, red, IR bands at disposal of a an area with a river, I want to discriminate water surfaces, do you have any good ideas on how to do that ? I did a NDWI which is not very convincing. NDVI neither. Multispectral sensor is Multispec4 from Airinov.

enter image description here

thermal:7.5-13.5 micrometer, sensor is a thermoMAP camera of Sensefly, produced by airinov I think

The NIR/(red+green+NIR) ratio is visible below, river streams are kind of visible but there is lot of noise: enter image description here

2

NDWI should be okay for it, but you can also look at NIR-ratio, which is the fraction of the total reflectance that is contributed by the NIR-band. In a case where you have the Red Edge band, I'd not count that and simply do:

NIR_ratio = NIR / ( Green + Red + NIR )

This approach gives a decent singleband ratio to look at. Shadows will consistently give you issues, regardless of which pixel-based method you use, so you could take it a few steps further and apply a object-oriented classification with a reasonable minimum object size, to remove isolated shadows.

  • Thanks for your answer @Mikkel , I tried different type of NIR ratio, still not very good, maybe it lies to the conditions: 1) the river I want to discriminate is quiet small in some places (50cm or less) and shallow. The resolution is about 12cm. 2) During drone data acquisition, sun insolation was variating (but sensor on top of drone can normaly correct that) – Raph Sep 17 '16 at 16:55
  • I will probably retry acquiring the data without shadows and apply some object oriented conditions to clean from noise. About that, do you know if acquisition before sun shining would be better ? or are other parameters bad during night for these bands? – Raph Sep 17 '16 at 17:02
  • @RaphaelVallat, you should aim at avoiding specular reflectance between the sun and the sensor, as this will make all delineation difficult. Furthermore, try to minimize the timeframe in which you acquire the imagery, to limit variability in sun angles and atmospheric variations. – Mikkel Lydholm Rasmussen Sep 17 '16 at 17:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.