Hi Jonathan,

Within eCognition you can use the "2D Feature Space Plot" which you fill find under Tools. To display your Ground Truth and your Classification at the same time, you could convert your Ground Truth into Samples. If you have the Ground Truth as Samples AND the Classification on one Image Object Level you can display both using the above mentioned tool. Samples will have a different marker (circle) in the 2D feature Space Plot compared to the classified image objects (cross marker).

If you already have classified image objects, this would be recommended workflow to get the Ground truth vector as Samples onto the same Level as your classification:

1. copy image object level: at Level 1: copy creating 'temp LVL' above [Create temp IOL to save the classification]
2. remove classification: at Level 1: remove classification [Delete Classification on original Level]
3. assign class by thematic layer: at Level 1: assign class by thematic layer using "ID" [Classify objects on the original Level based on the Ground Truth vector]
4. classified image objects to samples: tree at Analysis: classified image objects to samples [Convert the Ground Truth Objects to Samples]
5. remove classification: at Level 1: remove classification [Delete the classified objects on the original Level again. Those are already converted to Samples, so no worries ;)]
6. assign class: with Existence of super objects tree (1) = 1 at Level 1: tree [Get the classification from the temp level back to the original Level]
7. delete image object level: delete 'temp LVL' [Delete the temp level]

Now you have your Ground Truth as Samples and the Classification on the same Level. If you use the 2D Featur Space Plot now, you should see different markers for Samples (Ground Truth) and your Classification.

Hope this helps?

Cheers,
Matthias

Matthias,

It worked! Thank you very much, as this has been a road block for some time now.

I'm hoping you can provide a bit more information about what I produced and the ruleset you supplied above.

1. First, what exactly are samples? I understand that they represent the objects classified by the thematic layer (in my case ground truth vector points for trees), but what are samples typically used for in eCognition...mostly 2D feature space work or other applications too?
2. Is there a way to see a list of samples I have in a project if say I created multiple samples for different classified objects? When I ran the algorithm, I didn't see anything populate sample wise on the screen but assume there is a way to see a list.
3. In the screenshots below, I have the newly created 2D feature space plot for my project. My goal is to compare the locational coordinates of the ground truth trees (which are the red samples in the plot) to the classified objects I have for those trees (green crosses). The best way to do this comparison is by using the x and y center coordinates for my classified objects of interest in the "select feature for x/y axis" option box, correct?

Any information is appreciated. Thanks again!

Jon

Hi Jon,

Great to hear, you got it up and running.

1. Samples are typically used for Supervised classification and Accuracy Assessment. 
2. As far as I know its not possible to display samples in a list. But what you could check is the VIEW --> WINDOWS --> Image Object Information window tool. This allows you to display your objects in a list. You can sort them based on any feature and also filter by class. If you use the feature "Number of Overlapping thematic objects" based on the vector that you have you could easily navigate to those objects (the object table is linked to the view! If you click on an entry in the table it will highlight this object an center it in the view. Trick here is to zoom in quite a bit and then click on an object in the list.
3. I am not sure if I understand this last question. Using X and Y features for the X and Y axis will give you a scatterplot which depicts the location of the single objects in space. The plot is actually sort of interactive. If you click on an object the respective point in the scatter plot will be highlighted in red.

Hope this helps a bit?!

Cheers,

Matthias

Matthias,

Very helpful information! I didn't know the image object information window tool existed until now, but it seems quiet useful so far.

Thanks again for your assistance,

Jon