3

I was watching a video regarding image segmentation in ArcGIS Pro 2.4.0. It was an ESRI event (2018), so the presenters were ESRI employees. Here's the video I am referring to: https://www.youtube.com/watch?v=v40abIwGZWA

During the demonstration, they segmented an image using the Segmentation tool under the tab Imagery --> Classification Tools --> Segmentation. The image was a satellite image (I forget exactly which satellite it was from, but I think the spatial resolution was 50 cm/pixel). With this resolution, they said that the default settings for spatial and spectral detail were just fine to use. Below is an image with the default settings:

enter image description here

I am trying to perform image segmentation on very high-resolution drone imagery, with a spatial resolution of 1.1 cm/pixel.

If I can have higher spectral and spatial settings for my imagery than the defaults, then how would that affect my image classification?

2
  • What type of features are you interested in classifying? This will make a difference on the parameters.
    – Aaron
    Mar 25, 2020 at 16:07
  • Aaron, it’s a grain sorghum field, consisting of sorghum plants and palmar amaranth plants (weeds).
    – ihb
    Mar 26, 2020 at 17:01

1 Answer 1

3

The spectral and spatial detail values range from 0 to 20. Since they are using the default values of 15, you wouldn't be able to increase your spatial and spectral detail by much more. However, the small increases in values (even by 1) have shown difference in my 0.5 cm drone imagery. By increasing the spatial setting, the proximity of your objects to one another will be taken into account, so this is good if your objects of interest are small and close together. By increasing the spectral setting, small variations in the spectral values will have more of an influence. Increase your spectral settings if you have different objects that have similar spectral characteristics.

This explanation on segmentation in ArcGIS Pro may help: https://pro.arcgis.com/en/pro-app/help/analysis/image-analyst/segmentation.htm

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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